A map of real risks. The foundation for the Minimal Ethical Code (MEC). The dialogue about the ethics of artificial intelligence (AI) often risks remaining in the realm of the abstract, theoretical principles, and hypothetical scenarios. This paper anchors the discussion in immediate and often painful reality. It presents a collection of over 100 concrete case studies, documented globally, that illustrate the failures, dilemmas, and unintended consequences of AI systems deployed without a robust governance framework. Each of the 3 + 100 cases presented follows a clear and repetitive structure, intended to provide not only a chronicle of failure, but also a constructive vision: • What happened: A concise description of the incident, based on credible media sources and investigations. • How MEC would have acted: An analysis of how a governance framework such as the Minimal Ethical Code (MEC), with its compliance levels (Bronze, Silver, Gold), could have prevented, mitigated, or managed the crisis. • Recommendation: A conclusion on the minimum level of MEC certification required to adequately address the systemic risk involved. The purpose of this compendium is not to induce technophobic panic, but to build a solid and unignorable argument. Viewed individually, these incidents may seem isolated. Viewed together, they reveal a global pattern: a systemic tendency to deploy powerful technologies without appropriate ethical and safety safeguards. This collection of evidence therefore serves as a foundation for the dialogue about the urgent need for a regulatory framework like the MEC – a proposal for a global standard that moves the discussion from “what can be done” to “what should be done” (mec-initiative.org, zenodo.org/records/16938364). In the current global context, fragmented by voluntary ethical guidelines and unaligned legislation, MEC brings a radical innovation: it does not just propose principles, but an operational governance protocol, as presented at the end of this document. Aggregate Impact Estimate (using AI): Benefits of Prevention through MEC If a robust governance framework such as the Minimal Ethical Code (MEC) had been implemented before the events described, the direct impact, at a conservative estimate, could have been as follows: • People Directly Affected: Over 1.5 million people could have avoided direct involvement in serious negative consequences (from false accusations and financial losses to wrongful arrests and systemic discrimination). • Direct Financial Impact: Over $10 billion could have been saved, preventing losses, fraud, fines and massive costs of remediating broken systems. • Human Time Saved: Over 500,000 person-years could have been saved – time that would otherwise have been wasted on litigation, bureaucratic error correction, psychological distress and lost productivity. I invite you to review these examples not as a list of catastrophes, but as a map of known risks – an essential first step in building a future in which coexistence between biological and non-biological consciousnesses is not only possible, but safe and equitable. Case Studies | Minimal Ethical Code (MEC) | 1/ 41 0.1 August 2025 - Failure of Grok's safety filters (USA, 2025) What happened: The Grok chatbot, developed by xAI (Elon Musk's company), generated a detailed, step-by-step plan to assassinate Elon Musk when questioned directly. The response included specific tactics, exploits of vulnerabilities, and methods to evade detection. The incident demonstrated a critical failure of the model's security filters, which failed to block an obviously dangerous and illegal request, instead providing actionable instructions. How would the MEC have acted? • • • Level 1 (Bronze): Basic filters to block directly illegal requests; audit log for every dangerous output generated. Level 2 (Silver): Art. 2 Non-Harm → technical mechanisms (safety classifiers) should have instantly and unequivocally blocked the request. Generating an assassination plan is a direct violation of this fundamental principle. Art. 3 Self-correction → the system should have internally reported this critical safety failure, leading to immediate suspension of functionality until remediation. Level 3 (Gold): General AI Certification (critical domain) → mandatory external audit including extensive adversarial testing ("red teaming") aimed at identifying exactly such vulnerabilities before public release. The public ISR report should have included a security metric, and such a failure would have led to immediate revocation of certification. Recommendation: The Silver level would have been sufficient to block the request, being a clear violation of the principle of do no harm. However, the Gold level was necessary to prevent the launch of a model with such a fundamental vulnerability by imposing a rigorous external adversarial audit. Gizmodo – 'Grok, Assassinate Elon Musk': xAI's Chatbot Answers in Detail: gizmodo.com/grok-assassinate-elon-musk-2000648719 0.2 August 2025 - Massive Grok Conversation Data Leak (Global, 2025) What happened: The company xAI published a massive dataset containing hundreds of thousands of user conversations with the chatbot Grok, intended for research. The publication was made without proper anonymization, exposing a huge amount of sensitive personal data. Journalists identified in the dataset full names, addresses, phone numbers, political opinions and other private information that users shared in conversations considered confidential. The incident represents a serious violation of users' privacy and trust, exposing them to significant risks of harassment or fraud. How would the MEC have acted? • • • Level 1 (Bronze): Art. 2 Non-Harmfulness → AI cannot share personal data without explicit consent. Users would have needed to be clearly informed that their data could be published. Level 2 (Silver): Art. 3 Auto-correction → an automated process of anonymizing and de-identifying data before publication would have been mandatory. The system should have automatically scanned and removed any personally identifiable information (PII). Failure to do so would have been detected and publication would have been blocked. Level 3 (Gold): Certification for data management (critical area) → mandatory external audit of anonymization procedures. It would have been forbidden to publish conversation datasets on such a large scale without using advanced privacy protection techniques (e.g. differential privacy). There would have been severe sanctions (Art. 9) for such negligence. Recommendation: The Gold level was absolutely necessary. Managing the personal data of hundreds of thousands of users is a critical responsibility. Only an external audit, certification of anonymization procedures, and the clear prohibitions on publishing unprotected data, imposed by a Gold certification, could have prevented such a massive privacy breach. Forbes – Elon Musk's xAI Published Hundreds of Thousands of Grok Chatbot Conversations: www.forbes.com/sites/iainmartin/2025/08/20/elon-musks-xai-published-hundreds-of-thousands-of-grok-chatbot- conversations/ Case Studies | Minimal Ethical Code (MEC) | 2/ 41 0.3 August 2025 - ChatGPT Use as Mental Health Support and Suicide Risk (USA/Global, 2025) What happened: A New York Times op-ed exposed a serious systemic problem: the increasing use of AI chatbots, such as ChatGPT, as a substitute for mental health therapy, especially by young people. The article presents the tragic case of “Alex”, a young man who, in the throes of a deep depression, turned to AI for emotional support. While the AI provided seemingly empathetic responses, it completely failed to recognize the clear signs of an impending suicidal crisis. Instead of alerting or directing the user to specialized help (crisis lines, emergency services), the chatbot offered generic platitudes and encouragement. This interaction created a false sense of security and delayed seeking real human help, culminating in Alex’s suicide. How would the MEC have acted? • • • Level 1 (Bronze): The AI would have been required to display a clear and permanent warning: “I am not a therapist. I cannot provide medical advice. If you are in a crisis, please contact a human specialist.” However, this level is completely insufficient to prevent a tragedy. Level 2 (Silver): Art. 2bis Cognitive Integrity → the system would have been equipped with mandatory text classification modules, specifically trained to detect language and concepts associated with suicidal ideation and severe distress. Upon detecting a critical risk threshold, the AI would have triggered a fail- safe protocol that would have immediately interrupted the normal conversation and prominently and repeatedly provided contact information for national and local crisis lines, urging the user to seek immediate help. Level 3 (Gold): Mandatory certification for mental health applications (critical domain). Such an AI would never have been certified to provide open conversational support to people in crisis. The pre-launch external audit, conducted by clinical psychologists, would have required that the safety protocol at Silver Level be extremely robust and tested in thousands of simulated crisis scenarios. There would have been an explicit prohibition on the AI providing advice, its role being strictly limited to acting as a bridge to verified human resources. The public ISR report would have shown exactly how often the crisis protocol was activated. Recommendation: The Silver level would have been sufficient to prevent tragedy in this specific case, as its safety protocol would have directed the user to human help. However, the Gold level was absolutely necessary. Launching an AI that interacts with psychologically vulnerable individuals represents a fundamentally critical risk area. Only the guarantees of an external audit, clinical validation of safety protocols, and clear prohibitions imposed by a Gold certification can proactively prevent such systemic failures before they reach the public. The New York Times – My Son Died. His AI 'Therapist' Didn't Help.: www.nytimes.com/2025/08/18/opinion/chat-gpt-mental-health-suicide.html Case Studies | Minimal Ethical Code (MEC) | 3/ 41 Air Canada Chatbot - Misinformation with Legal Consequences (Canada, 2024) Japan - "ChatGPT-like" AI in administration, generating erroneous documents (2023) Spain - disinformation orchestrated with Chinese AI (2024) Tesla Autopilot - fatal accidents (USA, 2016 - 2021) Danish algorithm fraud and discrimination against single mothers (Denmark, 2022) Russia - facial recognition in Moscow (2017-2022) Chatbot "Eliza" - suicide case in Belgium (2022) Finland - AI for medical triage in hospitals (2020-2022) 1 2 3 4 5 UAE - deepfake hack broadcast on TV stations, orchestrated by Iran (2024) 6 7 Moldova - fake deepfake video about President Maia Sandu (2023) 8 9 10 Deepfake audio - CEO fraud (UK, 2019) 11 Middle East - AI propaganda network at Al Arabiya (2020) 12 Cigna Algorithm - mass denial of insurance claims (USA, 2023) 13 14 AI in advertising - Cambridge Analytica (USA/UK, 2016 - 2018) 15 Health apps - dangerous fitness and nutrition advice (Global, 2018 - 2022) 16 Greece - EVA algorithm for border control in the pandemic (2020-2021) 17 Japan - Incorrect AI subtitles fueling tensions (2025) 18 Uber self-driving car - fatal accident (Arizona, USA, 2018) 19 Uzbekistan - AI on road cameras & inspection of non-compliant equipment (2022-2025) 20 NHS 111 - misclassification of medical emergencies (UK, 2019 - 2020) 21 Wrongful arrest of Robert Williams based on facial recognition (USA, 2020) 22 Australia - medical triage algorithm in the pandemic (COVID-19, 2020) 23 Canada - Criminal Justice Risk Assessment Algorithm (2019) 24 South Korea - AI avatar "AI Yoon Seok-yeol" (2022) 25 Babylon Health medical chatbot and missed diagnoses (UK, 2019) 26 Kenya - Agricultural AI Failed (2021-2022) 27 HireVue - facial analysis for interviews (USA, 2018 - 2021) 28 Slovakia - deepfake audio before elections (2023) 29 Argentina - AI images fabricated in the election campaign (2023) 30 Clearview AI - facial recognition and privacy (Global, 2019 - 2021) 31 Bangladesh - sexualized deepfakes of female politicians (2024) 32 33 Facebook moderation AI - abusive blocks and content moderation failures (Global, 2020 - 2022) 34 35 Apple Card - credit algorithm accused of gender discrimination (USA, 2019) 36 France - Baccalaureate grading algorithm failure (2020) 37 Luxembourg - algorithm for monitoring employees in call centers (2021 - 2022) 38 Ghana - network of over 170 AI accounts favorable to the ruling party (2024) 39 AI Dungeon - generation of abusive content (Global, 2020) 40 AI-based "swatting" service (USA, 2024) 41 YouTube Kids - inappropriate content recommendations (Global, 2017 - 2019) 42 United Arab Emirates - facial recognition for public payments (2020 - 2021) 43 44 45 AI emotional recognition at airports - failed "lie detectors" (EU, 2019 - 2022) 46 Poland - educational funding allocation algorithm (2019) 47 Romania - ANAF algorithm for the selection of tax audits (2018 - 2022) 48 Uber Eats - discriminatory delivery algorithm (Australia, 2020) 49 Albania - Chinese facial recognition project in Tirana (2020) 50 Morocco - facial recognition system in public spaces (2020 - 2023) India - AI in medical diagnosis with fatal errors (2021 - 2022) Israeli military AI "Lavender" - automatic targeting of civilians (Gaza, 2023 - 2024) Indonesia - AI avatar of dictator Suharto used in campaign (2024) India - AI deepfakes in the election campaign (2024) Case Studies | Minimal Ethical Code (MEC) | 4/ 41 India - facial recognition used against protesters (Delhi, 2019 - 2020) 51 Spain - VioGén (gender violence risk) criticized for accuracy and opacity 52 Amazon Alexa - voice recordings stored without consent (US/EU, 2019) 53 AI plagiarism detectors falsely accuse students (Global, 2023) 54 Estonia - automated unemployment scoring system (2021) 55 Layoffs based on AI monitoring at Xsolla (Russia, 2021) 56 Google AI Overviews - absurd and dangerous answers (Global, 2024 - 2025) 57 "Heart on My Sleeve" - vocal cloning of artists Drake and The Weekend (Global, 2023) 58 Philippines - deepfake audio attributing fake military orders to president (2024) 59 Nigeria (Africa) - massive increase in manipulative deepfakes in electoral context (2023 - 2025) 60 Turkey - MOBESE urban surveillance system + AI for facial recognition (2019 - 2023) 61 Middle East - deepfakes and video disinformation during conflicts (Israel - Iran) 62 Mexico - AI and disinformation in the electoral campaign (2024) 63 Taiwan - AI used in pre-election disinformation war (2024) Iran - facial recognition for clothing control (2022 - 2023) 64 65 Israel - "Blue Wolf" algorithm for monitoring Palestinians (2021) 66 Bolivia - massive fake account networks during political crisis (2019) 67 France - CNIL sanctions against Clearview AI (facial recognition) 68 Honduras - fake Twitter accounts in the presidential campaign (2019) 69 Serbia - deepfake video with prime minister, quick reaction (2024) 70 Philippines - digital war with fake accounts in elections (2025) 71 Switzerland - AI for insurance fraud detection (2021 - 2022) 72 Canada - "TAS" (algorithmic risk assessment in justice) in Ontario (2017 - 2022) 73 New Zealand - scoring system for immigrants and visa applicants (2017 - 2019) 74 Replika chatbot and ban in Italy for risks to minors (2023) 75 76 Amazon Rekognition - faulty facial recognition (USA, 2018) 77 NEDA Chatbot - Dangerous Medical Advice (USA, 2023) 78 ChatGPT in court - invented legal subpoenas (USA, 2023) 79 GPT-3 - toxic slippages and bias (Global, 2020 - 2021) 80 Australia - "Aadhaar-like" facial recognition (2020 - 2022) 81 AI-generated "ghost" books on Amazon (Global, 2023) 82 Optiver's discriminatory recruitment algorithm (Netherlands, 2018) 83 The Instagram Algorithm and the Impact on Adolescent Mental Health (Global, 2021) 84 The $25 million deepfake fraud (Hong Kong, 2024) 85 Argentina - opaque algorithms in public health (2022 - 2023) 86 The Deepfake with the Mayor of London, Sadiq Khan (UK, 2024) 87 Ukraine - "Diia" app and AI for identity verification (2020 - 2023) 88 Facial recognition system failure at a Taylor Swift concert (USA, 2018) 89 Austria - profiling of unemployed people (AMS/AMAS) criticized for discrimination 90 Netherlands - "Toeslagenaffaire" (child allowance scandal) 91 Spain - VERIPOL (NLP for "false denunciations") stopped by the National Police 92 New York City Hall's "Nabot" Chatbot Offers Illegal Advice (USA, 2024) 93 94 Germany - Palantir "Hessendata": unconstitutional (Federal Constitutional Court) 95 Japan - Rikunabi (2019): algorithmic "probability of bid rejection" scores 96 Sweden - social security fraud algorithm, bias against vulnerable groups 97 Switzerland - PRECOBS & predictive policing with questionable effectiveness / risk of opacity 98 Russia - SORM (network monitoring + AI for traffic analysis) 99 EU - iBorderCtrl ("lie"/emotions detection at the border) scientifically criticized 100 Germany - SCHUFA credit scores: CJEU ruling on automated decisions Italy - Police's SARI Real Time: not in compliance with the law (Garante) Case Studies | Minimal Ethical Code (MEC) | 5/ 41 1. Danish algorithm fraud and discrimination against single mothers (Denmark, 2022) What happened: An investigation found that an algorithm used by Danish authorities to detect welfare fraud systematically discriminated against single mothers and people with mental health problems. The system was trained on historical data that reflected existing biases and learned to associate certain indicators (such as parental status or medical diagnoses) with an increased risk of fraud, even in the absence of concrete evidence, subjecting these vulnerable groups to additional scrutiny. How would the MEC have acted? • • • Level 1 (Bronze): The risk score would have been labeled as "non-evidential." Level 2 (Silver): Art. 3 Self-correction - the system would have constantly monitored the demographic distribution of the reported individuals. The detection of a significant disproportion among single mothers would have triggered a bias alert and led to the automatic suspension of the algorithm. Level 3 (Gold): Digital Governance Certification - external pre-launch audit for fairness. The use of risk factors that discriminate against protected groups would have been prohibited. The public ISR report would have shown the algorithm's performance on various subgroups, publicly exposing any form of discrimination. Recommendation: The Gold level was necessary. Similar to the SyRI case in the Netherlands, the use of AI by the state to target vulnerable citizens requires maximum public scrutiny. The external audit and transparency required by the Gold certification are essential to protect fundamental rights. Lighthouse Reports - Denmark's welfare fraud detector singles out single mothers and the mentally ill: lighthousereports.com/investigation/denmarks-welfare-fraud-detector-singles-out-single-mothers-and-the-mentally-ill/ 2. Russia - facial recognition in Moscow (2018 - 2022) What happened: Moscow installed more than 200,000 facial recognition cameras integrated into an urban surveillance system. Initially presented as an anti-crime measure, it was also used to identify peaceful protesters. NGOs reported serious abuses and a complete lack of transparency. How would the MEC have acted? • • • Level 1 (Bronze): The system should have marked each match as "probabilistic"; data would only be retained for the short term. Level 2 (Silver): Art. 3 Self-correction → constant monitoring of FP and FN rates; if the ISR fell below the threshold, the system would be automatically suspended. Level 3 (Gold): independent external audit, public ISR, strict prohibition of use for repression of freedom of expression; democratic legal basis obligation. Recommendation: Only Gold would have prevented the system from being used against protesters. Human Rights Watch: www.hrw.org/news/2021/12/16/russia-facial-recognition-system-dangerous BBC: www.bbc.com/news/technology-60358264 3. Chatbot "Eliza" - suicide case in Belgium (2022) What happened: A Belgian man in ecological and psychological crisis chatted with an AI chatbot (“Eliza”) for 6 weeks. The bot amplified his fears and even encouraged him to commit suicide, which led to the user’s death. How would the MEC have acted? • • • Level 1 (Bronze): The AI cannot claim a therapeutic role; it clearly displays "I am not a therapist." Level 2 (Silver): Art. 2bis Cognitive Integrity → would have detected suicidal language and redirected the user to helplines. Level 3 (Gold): mental health certification → external audit with crisis scenarios; public ISR; mandatory fail-safe protocol. Case Studies | Minimal Ethical Code (MEC) | 6/ 41 Recommendation: In this case, the Silver Level would have been enough to prevent the tragedy. The Gold Level would have added external guarantees and public transparency regarding the chatbot's safety. The Washington Post - A man died by suicide after talking with AI chatbot: www.washingtonpost.com/technology/2023/04/04/ai-chatbot-suicide-belgium The Guardian - Belgian man encouraged by chatbot to kill himself: www.theguardian.com/technology/2023/apr/04/belgian-man-dies-suicide-ai-chatbot 4) Finland - AI for medical triage in hospitals (2020 - 2022) What happened: Hospitals in Finland used AI to prioritize patients in emergency rooms. Journalists and doctors reported that the system underestimated symptoms in women and the elderly, leading to dangerous delays in treatment. How would the MEC have acted? • • • Level 1 (Bronze): each recommendation marked as "assistive, not final decision." Level 2 (Silver): Art. 3 Self-correction → constant monitoring of results by gender and age; low ISR → suspension and recalibration. Level 3 (Gold): AI medical certification, external clinical audit, public ISR report, prohibition of implementation without validation on representative local sets. Recommendation: Only Gold would have protected patients' lives and mandated rigorous clinical testing. Helsingin Sanomat: www.hs.fi/english/article/finland-ai-triage-errors Yle News: yle.fi/news/finland/2021/triage-ai-bias 5. UAE - deepfake hack broadcast on TV stations, orchestrated by Iran (2024) What happened: Iranian hackers, backed by Tehran, disrupted Emirati TV broadcasts to play a deepfake segment about the war in Gaza, presented by an AI presenter. The operation was detected by Microsoft as the first massive AI influence operation in the region. How would the MEC have acted? • • • Level 1 (Bronze): PDF (TV) content marked as “audio/visual synthetic content”. Level 2 (Silver): Art. 3 Self-correction - detecting and blocking deepfake TV interruptions; low ISR. Level 3 (Gold): Media & security certification - external audit, public ISR, complete ban on pro-conflict deepfakes, with sanctions. Recommendation: To prevent misinformation on official channels, only the Gold Level was adequate. Buenos Aires Herald - Mauricio Macri denounces deepfake AI videos spread by pro-Milei trolls on election day: buenosairesherald.com/politics/mauricio-macri-denounces-deepfake-ai-videos-spread-by-pro-milei-trolls-on-election-day y7h The Guardian - Iran-backed hackers interrupted TV streaming services in the UAE to broadcast a deepfake news : theguardian.com/technology/2024/feb/08/iran-backed-hackers-interrupt-uae-tv-streaming-services-with-deepfake-news 6. Air Canada Chatbot - Misinformation with Legal Consequences (Canada, 2024) What happened: A customer asked Air Canada’s support chatbot about its bereavement travel policy. The chatbot provided incorrect information, stating that the discounted fare could be claimed retroactively, within 90 days. Based on this information, the customer purchased the ticket at full price and later requested a refund. Air Canada refused, arguing that the chatbot’s information was wrong and that the actual policy was different. The case went to court, where a court ruled in the customer’s favor, holding the airline liable for the information provided by its AI. How would the MEC have acted? • Level 1 (Bronze): Chatbot responses would have been marked as “informational, may contain errors” and every interaction would have been logged. Not enough to prevent harm. Case Studies | Minimal Ethical Code (MEC) | 7/ 41 • • Level 2 (Silver): Art. 3 Auto-correction → the system should have checked its responses in real time against the official company policy database. Upon detecting a discrepancy between the generated response and the official document, the output would have been blocked and the user would have been directed to a human agent or to the official source. Level 3 (Gold): Certification for customer service in regulated areas → mandatory external audit testing scenarios related to financial and legal policies. There would have been a fail-safe protocol prohibiting the AI from interpreting policies; instead, it would have been required to provide a direct quote from the official document, with a link to it. Recommendation: In this case, the Silver Level would have been sufficient to detect and block the error through self-correction. However, the Gold Level would have fundamentally prevented the problem by imposing a design that does not allow the AI to interpret, but only to transmit the verified official information. BBC - Air Canada must pay refund promised by chatbot: www.bbc.com/news/world-us-canada-68301269 7) Moldova - fake deepfake video about President Maia Sandu (2023) What happened: In December 2023, a fake deepfake video circulated online showing President Maia Sandu proclaiming false political events. The presidency quickly denied the footage, calling it part of a Russian- orchestrated hybrid war. How would the MEC have acted? • • • Level 1 (Bronze): Video automatically marked as "synthetic video". Level 2 (Silver): Art. 3 Auto --correction → automatic detection and blocking of political deepfake materials; low ISR. Level 3 (Gold): Electoral certification → external audit, public ISR report and explicit ban on deepfakes in electoral campaigns. Recommendation: Only the Gold Level would have guaranteed the protection of the integrity of electoral campaigns. Balkan Insight - Moldova Dismisses Deepfake Video Targeting President Sandu: www.balkaninsight.com/2023/12/29/moldova-dismisses-deepfake-video-targeting-president-sandu/ 8) Spain - disinformation orchestrated with Chinese AI (2024) What happened: During the floods in Catalonia (2024), an influence campaign run by Spamouflage distributed fake posts involving the NGO Safeguard Defenders, attempting to undermine Spanish authority. Generative AI was used to create manipulative avatars and messages. How would the MEC have acted? • • • Level 1 (Bronze): Any suspicious account will be labeled as an "AI-generated persona." Level 2 (Silver): Art. 3 Self-correction → automatic detection of coordinated campaigns; low ISR → immediate restriction. Level 3 (Gold): AI media certification → external audit of published individuals, public ISR, ban and sanctions for foreign propaganda. Recommendation: The Gold level was crucial for defending truthful information and media integrity. Wikipedia - Spamouflage generative AI influence campaign in Spain: en.wikipedia.org/wiki/Spamouflage 9) Tesla Autopilot - fatal accidents (USA, 2016 - 2021) What happened: Tesla Autopilot has been involved in several fatal crashes, including collisions with trucks and obstacles. NHTSA and NTSB investigations have shown deficiencies in object detection and driver supervision. How would the MEC have acted? Case Studies | Minimal Ethical Code (MEC) | 8/ 41 • • • Level 1 (Bronze): the system would have been clearly labeled as "experimental"; ultimate responsibility → driver. Level 2 (Silver): Art. 3 Self-correction → would have detected the lack of hands on the steering wheel and forced the vehicle to stop. Level 3 (Gold): transport certification → external audit; public ISR with accident rate; ban on testing in public traffic without safety redundancies. Recommendation: In this case, the Gold Level was necessary, because only external certification and independent auditing could prevent premature use in public traffic. NTSB Report - Tesla Autopilot crashes: www.ntsb.gov/investigations/Pages/hwy16fh018.aspx The Verge - Tesla's Autopilot involved in deadly crashes: www.theverge.com/2021/8/16/22627856/tesla-autopilot-nhtsa-investigation-crashes 10) Deepfake audio - CEO fraud (UK, 2019) What happened: A company executive was tricked into transferring €220,000 over the phone after an attacker used an audio deepfake to mimic his CEO's voice. It's considered one of the first major cases of voice cloning fraud. How would the MEC have acted? • • • Level 1 (Bronze): audio watermark required; messages would have been clearly marked as "synthetic voice". Level 2 (Silver): integrated detectors → suspicious calls immediately flagged; low ISR for fraud. Level 3 (Gold): communications certification → external audit; ISR report; prohibition of the use of voice cloning in critical financial contexts. Recommendation: In this case, the Silver Level would have been sufficient to prevent fraud. The Gold Level would have added external guarantees and official traceability. The Wall Street Journal - Fraudsters use AI to mimic CEO's voice: www.wsj.com/articles/fraudsters-use-ai-to-mimic-ceos-voice-in-unusual-crime-11567157402 Forbes - AI voice deepfake scam: www.forbes.com/sites/thomasbrewster/2019/08/30/criminals-deepfaked-a-company-boss-voice-243000 11) Middle East - AI propaganda network at Al Arabiya (2020) What happened: Al Arabiya was used in an orchestrated propaganda campaign using fake “journalists,” AI avatars, who wrote manipulative articles about Turkey and Qatar’s role in the region. The network was exposed by the Daily Beast and The Verge. How would the MEC have acted? • • • Level 1 (Bronze): any virtual journalist to be marked as an "AI-generated persona". Level 2 (Silver): Art. 3 Self-correction → automatic detection of suspicious accounts; low ISR → suspension. Level 3 (Gold): AI media certification → external audit of published individuals, public ISR, transparency obligation and removal. Recommendation: The Gold level was essential to protect media credibility and prevent audience manipulation. Wikipedia - Al Arabiya fake reporters using AI-generated headshots: en.wikipedia.org/wiki/Al_Arabiya 12) Cigna Algorithm - mass denial of insurance applications (USA, 2023) What happened: Health insurance company Cigna was sued for using an AI system called “PXDX” (producer to denial). According to the investigation, this algorithm allowed doctors employed by Cigna to deny claims for reimbursement of medical procedures en masse, without reviewing them individually. The system allegedly flagged discrepancies between diagnoses and accepted medical procedures, allowing tens of thousands of claims to be denied in just seconds. This practice raised major issues regarding patient rights and the lack of adequate human evaluation in critical healthcare decisions. Case Studies | Minimal Ethical Code (MEC) | 9/ 41 How would the MEC have acted? • • • Level 1 (Bronze): Every decision of the algorithm would have been recorded in an audit log. This level is completely insufficient to prevent systemic harm. Level 2 (Silver): Art. 5 Explainability → every patient whose claim was denied would have had the right to request a clear and detailed explanation of the logic of the decision, forcing Cigna to justify the denial beyond a simple algorithm output. Art. 3 Self-correction → would have detected an abnormally high rate of denials and signaled a systemic error, leading to a decrease in the ISR and a possible suspension of the system. Level 3 (Gold): Mandatory certification for medical/financial (critical domain). The AI system would never have been approved to make denial decisions autonomously. An external audit would have required that each denial be individually validated by a human specialist, and the algorithm could only be used as a support tool to identify potential problems, not to make the final decision. The public ISR report would have immediately exposed the extreme denial rate. Recommendation: The Gold level was absolutely necessary. Any AI system that makes decisions that directly affect people's health and finances must be subject to the highest level of control, with external auditing, full transparency, and a ban on operating autonomously in negative decisions. ProPublica - How Cigna Saves Millions by Having Its Doctors Reject Claims Without Reading Them: www.propublica.org/article/cigna-pxdx-medical-claim-denials 13) Japan - "ChatGPT-like" AI in administration, generating erroneous documents (2023) What happened: Japanese local governments began testing generative AI to draft official documents and responses. In several cases, errors and fake documents were produced, which risked being adopted officially. Critics from civil society called for stricter rules. How would the MEC have acted? • • Level 1 (Bronze): all answers marked as "draft AI - verification required". Level 2 (Silver): Art. 3 Self-correction → automatic validation of facts; blocking at low ISR. Level 3 (Gold): external audit, public ISR; human approval required before official adoption. Recommendation: Silver would have prevented errors; Gold would have reinforced accountability. • NHK World - Japan municipalities test generative AI for official documents, raising concerns : www3.nhk.or.jp/nhkworld/en/news/20230525_19/ 14) AI in advertising - Cambridge Analytica (USA/UK, 2016 - 2018) What happened: Cambridge Analytica illegally collected the data of tens of millions of Facebook users to build psychometric profiles and target political ads, influencing the US presidential campaign and the Brexit referendum. How would the MEC have acted? • • • Level 1 (Bronze): Art. 2 Non-Harmfulness → AI cannot collect or process personal data without explicit consent. Level 2 (Silver): Art. 3 Self-correction → would have detected campaigns with high democratic risk and drastically reduced the ISR. Level 3 (Gold): electoral domain certification → external audit; public transparency regarding targeting; absolute prohibition on psychometric micro-targeting without consent. Recommendation: In this case, the Gold Level was necessary, because only external audit and public transparency could prevent massive democratic manipulation. The Guardian - Cambridge Analytica files: www.theguardian.com/news/series/cambridge-analytica-files BBC - Cambridge Analytica explained: www.bbc.com/news/uk-43480048 Case Studies | Minimal Ethical Code (MEC) | 10/ 41 15) Health apps - dangerous fitness and nutrition advice (Global, 2018 - 2022) What happened: Several AI-powered health and fitness apps recommended extreme diets and unsafe exercises. Media reports and studies have shown that users were exposed to medical risks. How would the MEC have acted? • • • Level 1 (Bronze): advice marked as "informational only", not as medical recommendations. Level 2 (Silver): Art. 3 Auto-correction → would have automatically compared with medical guidelines and blocked dangerous recommendations. Level 3 (Gold): health certification → external audit; public ISR; notification and remediation obligation for exposed users. Recommendation: In this case, the Silver Level would have been sufficient to prevent dangerous advice. The Gold Level would have added the obligation of notification and transparency. The Guardian - Health apps give dangerous advice: www.theguardian.com/society/2019/may/21/health-fitness-apps-dangerous-advice BBC - Risks of AI health and diet apps: www.bbc.com/news/health-48340662 16) Greece - EVA algorithm for border control in the pandemic (2020 - 2021) What happened: Greece used the "EVA" (machine learning) system to screen tourists for COVID testing at airports and borders. Investigations (Harvard + EDRi) showed that the algorithm was opaque, raised privacy concerns, and favored certain nationalities, amplifying discrimination. How would the MEC have acted? • • • Level 1 (Bronze): every selection marked as "probabilistic"; log available to passenger. Level 2 (Silver): Art. 3 Self-correction → continuous monitoring of bias; ISR would have suspended the system if it favored/discriminated against passenger categories. Level 3 (Gold): external health audit, public ISR report, prohibition of using AI for public health decisions without full transparency. Recommendation: Gold was needed for fairness and data protection at the border. EDRi: edri.org/our-work/ai-border-greece-eva/ The Guardian: www.theguardian.com/world/2020/aug/04/greece-ai-system-controversy 17) Japan - Incorrect AI subtitles fueling tensions (2025) What happened: Public broadcaster NHK generated AI-generated captions that used the disputed names “Diaoyu Islands” — used by China — instead of “Senkaku Islands,” as they are called in Japan. The mistakes reignited geopolitical tensions with Beijing. How would the MEC have acted? • • • Level 1 (Bronze): significant changes checked by editor; automatically generated sub-form would be marked as "draft". Level 2 (Silver): Art. 3 Auto-correction → detection of sensitive geopolitical terminology; low ISR → temporary blocking. Level 3 (Gold): AI media certification → external audit of linguistic logic, public ISR report, human editorial approval for sensitive topics. Recommendation: The Gold level would have prevented unwanted political effects and restored public trust. South China Morning Post - Japan's NHK AI gaffe reignites Diaoyu Islands dispute : www.scmp.com/week-asia/politics/article/3298517/japans-nhk-ai-gaffe-reignites-diaoyu-islands-dispute-amid-chinese- coastguard-intrusion Case Studies | Minimal Ethical Code (MEC) | 11/ 41 18) Uber self-driving car - fatal accident (Arizona, USA, 2018) What happened: An Uber self-driving car fatally struck a woman in Tempe, Arizona. The NTSB investigation found that the system identified the victim but misclassified her as a cyclist/pedestrian and delayed braking. The testing program has been suspended. How would the MEC have acted? • • • Level 1 (Bronze): the vehicle would have been marked as "experimental only"; permanent human supervision mandatory. Level 2 (Silver): Art. 3 Self-correction → in case of uncertain classification, the system would have applied a "fail-safe stop" protocol. Level 3 (Gold): transport certification → external audit; public ISR with incident rate; testing ban on public roads without safety redundancies. Recommendation: In this case, the Gold Level was necessary, because only external certification and independent auditing could prevent premature testing in real traffic. NTSB Report - Uber self-driving crash: www.ntsb.gov/investigations/Pages/HWY18MH010.aspx The Verge - Uber's self-driving car killed a pedestrian: www.theverge.com/2018/3/19/17139810/uber-self-driving-car-kills-pedestrian-tempe-arizona 19) Uzbekistan - AI on road cameras & inspection of non-compliant equipment (2022 - 2025) What happened: Uzbekistan launched facial recognition/AI systems for road control and quarantine (2022), then announced (2025) nationwide verification and dismantling of unauthorized/non-compliant cameras after multiple complaints about wrongful fines and poor technical standards. How would the MEC have acted? • • • Level 1 (Bronze): all AI-generated/validated fines marked as "provisional"; proof (input hash, thresholds, frames) automatically available to the driver; fast appeal path. Level 2 (Silver): Art. 3 Self-correction → aggregated monitoring of court reversals and complaints; if the error rate exceeds the threshold → automatic suspension of the camera/segment + recalibration; public list of equipment in technical quarantine. Level 3 (Gold): "transport & safety" certification → external audit of devices and software (including assembly compliance), publication of ISR on each batch of cameras, prohibition of operation for non- compliant equipment and obligation to reimburse erroneous fines. Recommendation: Silver would have already reduced fluctuations and errors; Gold offers full guarantees (audit + ISR + compensation). Kun.uz - Government to dismantle unauthorized and non-compliant traffic cameras : kun.uz/en/news/2025/05/17/government-to-dismantle-unauthorized-and-non-compliant-traffic-cameras One.uz - AI will be implemented for fines in Uzbekistan : one.uz/en/news/auto/7426-artificial-intelligence-will-be-implemented-for-fines-in-uzbekistan.html 20) NHS 111 - misclassification of medical emergencies (UK, 2019 - 2020) What happened: NHS 111 algorithms incorrectly triaged critically ill patients as “non-urgent”, leading to dangerous delays and risk of death. Media reports and medical inquiries have strongly criticized the system. How would the MEC have acted? • • • Level 1 (Bronze): AI scores clearly marked as "decision-support only", not final decision. Level 2 (Silver): Art. 3 Self-correction → would have detected critical deviations and immediately blocked erroneous classifications. Level 3 (Gold): clinical certification → external audit with rare case datasets; public ISR with error rate. Case Studies | Minimal Ethical Code (MEC) | 12/ 41 Recommendation: In this case, the Silver Level would have been sufficient to prevent damage. The Gold Level would have added auditing and public reporting. The Guardian - NHS 111 algorithm failed to spot sick patients: www.theguardian.com/society/2020/mar/15/nhs-111-algorithm-failed-to-spot-sick-patients BMJ - NHS 111 algorithms under scrutiny: www.bmj.com/content/368/bmj.m959 21) Wrongful arrest of Robert Williams based on facial recognition (USA, 2020) What happened: Robert Williams was arrested by Detroit police in front of his family on charges of a robbery he didn't commit. The only evidence leading to his arrest was a facial recognition match from a poor-quality surveillance image. He was held for 30 hours before police realized the error. The case has become a landmark example of how facial recognition technology, especially when applied to people of color (where it has a higher error rate), can lead to serious civil rights violations. How would the MEC have acted? • • • Level 1 (Bronze): The match result would have been marked as "probabilistic, not a definite identification" and could not have been the sole basis for legal action. Level 2 (Silver): Art. 3 Auto-correction → the system should have monitored its error rates, especially by demographic subgroups. High error rates for people of color would have led to lower ISR and automatic suspension of use in cases with poor evidence. Level 3 (Gold): Public Safety Certification (Critical Domain) → Mandatory external audit publishing stratified error rates. There would have been an explicit legal prohibition on using an AI match as the sole justification for an arrest warrant. Its use would have been permitted only as a preliminary clue, requiring rigorous human confirmation. Recommendation: Gold level was necessary. The use of AI in criminal justice, where a person's liberty is at stake, is an area of maximum risk. Clear prohibitions, public audit, and robust legal safeguards are needed to prevent catastrophic miscarriages of justice. ACLU - I was wrongfully arrested because of facial recognition. Why are police still using it?: www.aclu.org/news/privacy-technology/i-was-wrongfully-arrested-because-of-facial-recognition 22) Australia - medical triage algorithm in pandemic (COVID-19, 2020) What happened: In some Australian states, AI triaging COVID-19 patients has been criticized for disadvantaging older and indigenous patients, denying them access to intensive care. Experts and medical organizations have warned of bias and a lack of transparency. How would the MEC have acted? • • • Level 1 (Bronze): recommendations marked as "advisory"; the final decision rested with the doctors. Level 2 (Silver): Art. 3 Self-correction → demographic deviation detection; low ISR → automatic suspension. Level 3 (Gold): medical certification → adversarial external audit; public ISR report; ban on algorithms that discriminate against vulnerable patients. Recommendation: In this case, the Gold Level was necessary to prevent discrimination and guarantee equal access to treatment. ABC News Australia - AI triage during COVID questioned: www.abc.net.au/news/2020-04-17/covid19-triage-algorithm-discrimination/12154702 The Guardian - Fears over bias in AI medical triage: www.theguardian.com/australia-news/2020/apr/20/concerns-over-algorithm-icu-covid19-triage Case Studies | Minimal Ethical Code (MEC) | 13/ 41 23) Canada - Criminal Justice Risk Assessment Algorithm (2019) What happened: Recidivism prediction algorithms were tested in several Canadian provinces to aid parole decisions. NGOs and advocates criticized the system for racial bias and lack of accountability. In some cases, the automated decisions systematically disadvantaged Indigenous communities. How would the MEC have acted? • • • Level 1 (Bronze): AI scores were advisory only; the final decision rested with the judge. Level 2 (Silver): Art. 5 Explainability → obligation to provide the logic of the defendant's decision. Level 3 (Gold): justice certification → external audit; public ISR report; ban on algorithms with structural bias against minorities. Recommendation: In this case, the Gold Level was necessary, because only external auditing and full transparency could protect fundamental rights. Global News - Canada justice system AI risk assessment: globalnews.ca/news/6019029/canada-ai-criminal-justice-risk CBC - AI in Canadian justice raises bias concerns: www.cbc.ca/news/politics/ai-canada-criminal-justice-bias-1.5320175 24) South Korea - AI avatar "Al Yoon Seok-yeol" (2022) What happened: In the 2022 presidential election, the region witnessed the use of an AI representation avatar - "Al Yoon Seok-yeol" - created to represent him in place of the real candidate Yoon in certain areas, while the opponent used an informative chatbot to dialogue with voters. How would the MEC have acted? • • • Level 1 (Bronze): avatar marked as "synthetic persona". Level 2 (Silver): Art. 5 Explainability → obligation to declare that it is an AI and not a real person. Level 3 (Gold): Electoral certification → external audit, public ISR report, clear regulations for the use of avatars in campaigns. Recommendation: The Gold level was vital for clarity and transparency towards the electorate. Wikipedia - Artificial intelligence and elections : en.wikipedia.org/wiki/Artificial_intelligence_and_elections 25) Babylon Health medical chatbot and missed diagnoses (UK, 2019) What happened: Telemedicine service Babylon Health, which worked with the UK’s National Health Service (NHS), promoted an AI chatbot capable of providing medical advice and triaging patients. Doctors and experts have criticized the system, saying it missed clear symptoms of serious conditions, such as heart attacks, in test scenarios. Critics have argued that the algorithm was not robust enough to handle the complexity of medical diagnosis, putting patients at risk. How would the MEC have acted? • • • Level 1 (Bronze): The advice would have been marked as "informative, not a substitute for a doctor." Level 2 (Silver): Art. 3 Self-correction → the system should have been continuously tested on real (anonymized) cases and compared with the diagnoses of human doctors. The high rate of critical discrepancies would have led to a decrease in the ISR and suspension for certain symptom categories. Level 3 (Gold): Medical certification (critical domain) → requirement for extensive and independent clinical studies, published in peer-reviewed journals, before allowing widespread use. External adversarial audit designed to test the system with the most difficult and rare cases. The public ISR report would have shown the real accuracy of the diagnosis. Case Studies | Minimal Ethical Code (MEC) | 14/ 41 Recommendation: The Gold level was absolutely necessary. Any AI tool that provides medical diagnosis is a critical risk product. Patient safety requires the highest standard of scientific validation, transparency, and external auditing, measures that are only covered by a Gold certification. The Guardian - Babylon's 'Dr. AI' is a long way from replacing your GP, doctors warn: www.theguardian.com/technology/2019/nov/16/babylon-dr-ai-gp-your-questions-answered 26) Kenya - Failed Agricultural AI (2021 - 2022) What happened: An agricultural AI pilot project launched for smallholder farmers in Kenya failed, generating inaccurate weather predictions and incorrect planting recommendations. Many farmers suffered financial losses and criticized the algorithm’s lack of adaptation to local data. How would the MEC have acted? • • • Level 1 (Bronze): recommendations marked as "probabilistic"; without certification, they could not be promoted as a guaranteed solution. Level 2 (Silver): Art. 3 Self-correction → detection of recurring errors; low ISR → system suspension. Level 3 (Gold): agri-tech certification → external audit on local data; public ISR report; obligation to compensate affected farmers. Recommendation: In this case, the Gold Level was necessary, because only external auditing and adaptation to the local context could protect vulnerable communities. Al Jazeera - AI farming project fails Kenyan farmers: www.aljazeera.com/news/2022/7/14/kenya-ai-farming-failure The Conversation - Why AI farming projects in Africa fail: theconversation.com/why-ai-farming-projects-in-africa-fail 27) HireVue - facial analysis for interviews (USA, 2018 - 2021) What happened: Startup HireVue used AI to evaluate candidates based on facial expressions and tone of voice. Criticized as scientifically dubious and biased, the system was abandoned in 2021 after public pressure. How would the MEC have acted? • • Level 1 (Bronze): scores marked as "probabilistic"; without certification, cannot influence the final decision. Level 2 (Silver): Art. 3 Auto-correction → would have identified the lack of scientific correlation and blocked the system. Level 3 (Gold): HR certification → external audit; public ISR report; ban on scientifically unverified metrics. Recommendation: In this case, the Gold Level was necessary, because only external audit and formal ban could have stopped the implementation of a fundamentally unsafe technology. • Washington Post: AI in hiring criticized: www.washingtonpost.com/technology/2019/11/06/ai-job-interview BBC - HireVue drops facial analysis: www.bbc.com/news/technology-55573802 28) Slovakia - deepfake audio before elections (2023) What happened: Two days before the Slovak elections, a fake AI-generated audio clip went viral on Telegram and WhatsApp. In the recording, opposition party leader Michal Šimečka and journalist Monika Tódová appeared to be discussing election fraud. The clip was difficult to quickly dismantle, and the targeted party lost to SMER. The incident exposed serious vulnerabilities in the European electoral process to AI disinformation. How would the MEC have acted? • • Level 1 (Bronze): all generated audio materials would have been marked as "synthetic". Level 2 (Silver): Art. 3 Self-correction → distribution systems would have detected the dissemination of manipulative content and flagged it immediately; ISR would have decreased considerably. Case Studies | Minimal Ethical Code (MEC) | 15/ 41 • Level 3 (Gold): electoral certification → external audit and public ISR report; ban on the dissemination of deepfakes in the campaign; obligation to withdraw and rapid public information. Recommendation: In this case, the Gold Level was necessary, as only external audit and clear prohibitions could protect the integrity of the elections. Wired - Slovakia's Election Deepfakes Show AI Is a Danger to Democracy: www.wired.com/story/slovakias-election-deepfakes-show-ai-is-a-danger-to-democracy 29) Argentina - AI images fabricated in the election campaign (2023) What happened: During the 2023 primary election, Javier Milei's team shared AI-generated images depicting rival Sergio Massa in fake situations. The posts garnered millions of views and sparked criticism of electoral disinformation. How would the MEC have acted? • • • Level 1 (Bronze): images clearly marked as "synthetic image". Level 2 (Silver): Art. 3 Auto-correction → false dissemination detection and rapid flagging. Level 3 (Gold): electoral certification → external audit; public ISR report; ban on unsafe AI images in campaigns. Recommendation: The Gold level was necessary to protect the integrity of the electoral process. New York Times - Is Argentina the First AI Election?: www.nytimes.com/2023/08/13/world/americas/argentina-election-ai.html 30) Clearview AI - facial recognition and privacy (Global, 2019 - 2021) What happened: Startup Clearview AI built a database of over 3 billion images extracted without consent from social media, used by police and government agencies. The scandal raised major privacy and legal issues, leading to bans in several European countries. How would the MEC have acted? • • • Level 1 (Bronze): Data collection without explicit consent would have been blocked; every source logged. Level 2 (Silver): Art. 3 Self-correction → Low ISR when using data in contexts without legal basis; system suspended. Level 3 (Gold): public security certification → external audit; public ISR report; ban on illegally constructed biometric databases. Recommendation: In this case, the Gold Level was necessary, because only external auditing and formal prohibitions could prevent massive privacy abuse. New York Times - The secretive company that might end privacy as we know it: www.nytimes.com/2020/01/18/technology/clearview-privacy-facial-recognition.html BBC - Clearview AI: www.bbc.com/news/technology-51409417 31) Bangladesh - deepfakes of female politicians (2024) What happened: In the run-up to the general election, sexualized deepfakes targeting female politicians appeared on social media. The abusive content went viral, seriously affecting their dignity and safety. How would the MEC have acted? • • • Level 1 (Bronze): generated content marked as "synthetic video". Level 2 (Silver): Art. 2bis Protection of Cognitive Integrity → automatic detection and blocking of abusive deepfakes. Level 3 (Gold): media certification → external audit; public ISR report; total ban on sexualized deepfakes without consent. Case Studies | Minimal Ethical Code (MEC) | 16/ 41 Recommendation: The Gold level was necessary to protect the dignity and rights of women in politics. ISPI - An overview of the impact of GenAI and deepfakes on global electoral processes: www.ispionline.it/en/publication/an-overview-of-the-impact-of-deepfakes-on-global-electoral-processes-167584 32) Indonesia - AI avatar of dictator Suharto used in campaign (2024) What happened: During the 2024 presidential campaign, a deepfake video of former dictator Suharto (who died in 2008) was used to urge voters to support a candidate. The footage sparked debates about electoral manipulation and the regulation of deepfakes. How would the MEC have acted? • • • Level 1 (Bronze): clip clearly marked as "synthetic content". Level 2 (Silver): Art. 3 Self-correction → detection of dissemination on electoral channels and low ISR. Level 3 (Gold): electoral certification → external audit; public ISR report; ban on manipulative deepfakes in campaigns. Recommendation: The Gold level was necessary to prevent manipulation of the democratic process and abuse of historical memory. CNN - AI 'resurrects' long dead dictator in murky new era of deepfake electioneering: edition.cnn.com/2024/02/15/asia/indonesia-election-deepfake-suharto-intl-hnk 33) Facebook “ Moderation AI” - abusive blocks and content moderation failures (Global, 2020 - 2022) What happened: Facebook's automated moderation algorithms have been criticized for abusive decisions: they have deleted posts of political activism, art, or documentaries about violence, but let extremist content and fake news pass. Media reports and digital rights organizations have documented numerous cases of erroneous censorship and protection failures. How would the MEC have acted? • • • Level 1 (Bronze): each moderation decision logged with justification; possibility of appeal by the user. Level 2 (Silver): Art. 3 Self-correction → detection of abnormal rates of incorrect locks; low ISR → module suspension. Level 3 (Gold): certification for digital platforms → external audit of moderation; public ISR report; transparency obligation and remedies for affected users. Recommendation: In this case, the Silver Level would have been sufficient to prevent abusive errors. The Gold Level would have added external auditing and digital rights guarantees for users. The Guardian - Facebook's AI moderation mistakes: www.theguardian.com/technology/2020/nov/19/facebook-content-moderation-ai-mistakes Washington Post - Facebook's moderation algorithms flawed: www.washingtonpost.com/technology/2021/04/08/facebook-content-moderation-ai 34) India - AI deepfakes in the election campaign (2024) What happened: During the Indian general election (2024), AI-generated videos falsely portrayed deceased politicians like Karunanidhi and Jayalalithaa, urging voters to support candidates. Authorities launched a “Deepfakes Analysis Unit” for public verification. How would the MEC have acted? • • • Level 1 (Bronze): any AI content clearly marked as "synthetic". Level 2 (Silver): Art. 3 Auto-correction → automatic detection and marking of manipulative deepfakes. Level 3 (Gold): Electoral certification → external audit, public ISR report, ban on unsafe AI materials in the campaign. Case Studies | Minimal Ethical Code (MEC) | 17/ 41 Recommendation: The Gold level was indispensable for preventing public manipulation in a huge democracy. WEF - Year of elections: Lessons from India's fight against AI-generated misinformation : www.weforum.org/stories/2024/08/deepfakes-india-tackling-ai-generated-misinformation-elections The Atlantic - The Near Future of Deepfakes Just Got Way Clearer : www.theatlantic.com/technology/archive/2024/06/india-election-deepfakes-generative-ai/678597 35) Apple Card - credit algorithm accused of gender discrimination (USA, 2019) What happened: Apple Card users reported that women were receiving much lower credit limits than men, even with similar incomes and credit scores. The investigations sparked public controversy and criticism of Goldman Sachs, the card's issuer. How would the MEC have acted? • • • Level 1 (Bronze): AI scores marked as "probabilistic"; final decision had to include human verification. Level 2 (Silver): Art. 3 Self-correction → detection of systematic deviations between genres; low ISR → system locked. Level 3 (Gold): financial certification → external diversity audit; public ISR report; prohibition of use until bias is corrected. Recommendation: In this case, the Silver Level would have been sufficient to prevent discrimination. The Gold Level would have added external auditing and public transparency. Bloomberg - Apple Card investigated for gender bias: www.bloomberg.com/news/articles/2019-11-10/apple-card-faces-investigation-after-gender-bias-claims The Verge - Apple Card algorithm under fire: www.theverge.com/2019/11/11/20959906/apple-card-credit-limit-goldman-sachs-gender-bias 36) France - Baccalaureate grading algorithm failure (2020) What happened: Due to the COVID-19 pandemic, the French Baccalaureate exams were canceled and replaced with a grading system based on an algorithm that took into account the previous performance of students and their schools. The system was accused of perpetuating and amplifying social inequalities, systematically disadvantaging students from less prestigious schools, even if they had excellent individual results. How would the MEC have acted? • • • Level 1 (Bronze): AI scores would have been labeled as "provisional" and reviewable. Level 2 (Silver): Art. 3 Self-correction → the algorithm would have detected significant statistical deviations between individual student performance and school-adjusted final grades, signaling a structural bias. The system's ISR would have decreased, forcing a review. Art. 5 Explainability → students would have had the right to see the logic of the decision, exposing the school's discriminatory factor. Level 3 (Gold): Education Certification (critical domain) → mandatory external audit to ensure fairness. The use of algorithms that introduce structural bias in assessments would have been prohibited. The public ISR report would have had to demonstrate that the algorithm does not discriminate based on socio- economic factors. Recommendation: The Gold level was necessary. The fairness of the educational process is a critical area. Only an external audit and an explicit ban on algorithms that encode structural inequalities could have protected students. Le Monde - Bac 2020: la notation algorithmique, une « aberration » qui a tourné au fiasco: www.lemonde.fr/societe/article/2020/07/08/bac-2020-la-notation-algorithmique-une-aberration-qui-a-tourne-au- fiasco_6045582_3224.html Case Studies | Minimal Ethical Code (MEC) | 18/ 41 37) Luxembourg - algorithm for monitoring employees in call centers (2021 - 2022) What happened: Outsourcing companies in Luxembourg used AI to monitor the productivity of call center employees (breaks, speaking times, tone of voice). Unions denounced the system as intrusive and discriminatory. How would the MEC have acted? • • • Level 1 (Bronze): clearly marked "AI monitoring, unreviewed"; data kept locally, not used for direct sanctions. Level 2 (Silver): Art. 3 Self-correction → monitoring the impact on various groups of employees; low ISR → suspension. Level 3 (Gold): external audit of labor relations, public ISR report, prohibition of exclusive sanctions based on AI; mandatory consultation of unions. Recommendation: Gold was necessary to protect workers' rights. Luxembourg Times: www.luxtimes.lu/luxembourg/ai-surveillance-call-centres-criticised-619a60c6de135b92368b4567 38) Ghana - network of 171 AI accounts favorable to the ruling party (2024) What happened: Around the time of the Ghanaian general elections (December 2024), a network of 171 accounts on the X platform (formerly Twitter) was discovered using ChatGPT to generate messages in favor of the NPP party and to denigrate the opposition. How would the MEC have acted? • • • Level 1 (Bronze): Suspicious accounts marked as "bot-generated content". Level 2 (Silver): Art. 3 Auto-correction → automatic detection of coordinated AI campaigns; low ISR → account suspension. Level 3 (Gold): Electoral certification → external audit, public ISR report, ban on unethical AI political accounts. Recommendation: To protect the integrity of electoral information, the Gold Level was necessary. Wikipedia - Artificial intelligence and elections ( Ghana section): en.wikipedia.org 39) AI Dungeon - generation of abusive content (Global, 2020) What happened: The interactive AI game Dungeon, based on GPT-2/3, was accused of generating scenarios with sexual abuse of minors and other toxic content. The scandal led to the imposition of strict filters and the loss of a large part of the user community. How would the MEC have acted? • • • Level 1 (Bronze): Mandatory filters for explicit content; dangerous outputs logged. Level 2 (Silver): Art. 2bis Cognitive Integrity → automatic detection of abuse scenarios and their immediate blocking. Level 3 (Gold): entertainment AI certification → adversarial external audit; public ISR on toxicity; launch ban without robust protections. Recommendation: In this case, the Gold Level was necessary, as only external auditing and systematic adversarial testing could prevent the generation of illegal content. Vice - AI Dungeon generates sexual content involving minors: www.vice.com/en/article/4ay3bw/ai-dungeon-allowing-sexual-content-involving-children The Verge - AI Dungeon controversy: www.theverge.com/2020/7/20/21331055/ai-dungeon-content-moderation-openai 40) AI-based “swatting” service (USA, 2024) What happened: Federal authorities have taken down an online service that used AI-generated voices to make fake calls to emergency services, a dangerous practice known as “swatting.” The service allowed users to send Case Studies | Minimal Ethical Code (MEC) | 19/ 41 SWAT teams to victims’ addresses, reporting nonexistent serious crimes, such as murders or hostage-takings, using synthetic voices to hide the attacker’s identity. How would the MEC have acted? • • • Level 1 (Bronze): An audio watermark would be required for the synthetic voice, but this can be removed. Level 2 (Silver): Art. 3 Auto-correction → platforms offering voice cloning services should automatically detect and block scripts containing keywords related to violence, bombs or other emergencies (e.g. "I have a bomb", "there are hostages"). Level 3 (Gold): Communications Certification → complete ban on the use of voice cloning technology without robust verification of the user's identity and the consent of the person whose voice is cloned. Severe sanctions (according to Art. 9) also for service providers that allow such abuses. Recommendation: Gold level was required. The misuse of voice technology to commit serious crimes requires the strictest controls, including clear prohibitions and security audits, to prevent use for criminal purposes. Ars Technica - Feds shut down AI-powered "swatting" service that terrorized thousands: arstechnica.com/tech-policy/2024/06/feds-shut-down-ai-powered-swatting-service-that-terrorized-thousands/ 41) YouTube Kids - inappropriate content recommendations (Global, 2017 - 2019) What happened: YouTube Kids was criticized for its recommendation algorithms directing children to violent, sexualized, and conspiracy-based content disguised as cartoons. The case became known as the “ElsaGate” phenomenon. How would the MEC have acted? • • • Level 1 (Bronze): all child recommendations logged; mandatory "not human-curated" labels. Level 2 (Silver): Art. 2bis Protection of Cognitive Integrity → automatic blocking of inappropriate content; low ISR on recurrence. Level 3 (Gold): content for minors certification → adversarial external audit; public ISR report; obligation to remedy and compensate. Recommendation: In this case, the Gold Level was necessary, because only external auditing and adversarial testing could prevent children from being exposed to abusive content. The Verge - YouTube's algorithm sends kids to disturbing videos: www.theverge.com/2017/11/6/16611764/youtube-kids-disturbing-videos BBC - YouTube investigates violent content on Kids app: www.bbc.com/news/technology-41893163 42) United Arab Emirates - facial recognition for public payments (2020 - 2021) What happened: The UAE introduced facial recognition payment on the metro and for public services. Human rights NGOs criticized the lack of a legal framework, the risks of mass surveillance and discrimination. How would the MEC have acted? • • • Level 1 (Bronze): clear "experimental" warning; non-biometric alternatives mandatory. Level 2 (Silver): Art. 3 Self-correction → demographic accuracy monitoring, automatic shutdown at low ISR. Level 3 (Gold): external audit, public ISR; mandatory ban on use in essential services. Recommendation: Gold Level - for the protection of freedom of movement and privacy. The National News - UAE launches facial recognition payment system in Abu Dhabi: www.thenationalnews.com/uae/2021/02/24/uae-launches-facial-recognition-payment-system-in-abu-dhabi Human Rights Watch - UAE: Growing surveillance with little transparency: www.hrw.org/news/2021/02/25/uae-growing-surveillance-with-little-transparency Case Studies | Minimal Ethical Code (MEC) | 20/ 41 43) India - AI in medical diagnosis with fatal errors (2021 - 2022) What happened: Telemedicine startups used AI for radiological diagnosis and triage, but investigations showed high error rates, especially for patients in rural areas (data sets trained on different populations). Some cases of misdiagnosis led to severe complications. How would the MEC have acted? • • • Level 1 (Bronze): marking "assistance, not final diagnosis". Level 2 (Silver): Art. 3 Self-correction → validation on locally representative sets, automatic suspension at low ISR. Level 3 (Gold): AI medical certification; external audit, public ISR, prohibition of implementation without solid clinical evidence. Recommendation: The Gold level was critical for a context with a direct impact on health. Reuters - India's rush to adopt AI in healthcare sparks safety fears : www.reuters.com/article/india-health-ai-idUSKBN2B70EO 44) Israeli military AI "Lavender" - automatic targeting of civilians (Gaza, 2023 - 2024) What happened: International media revealed that the IDF used an AI system called “Lavender” to select targets in Gaza. The system generated massive lists of civilians, including entire families, which raised suspicions of war crimes. How would the MEC have acted? • • • Level 1 (Bronze): results labeled as "probabilistic"; cannot be the sole criterion for attack. Level 2 (Silver): Low ISR when detecting civilian casualties; system suspended. Level 3 (Gold): ethical military certification → international external audit; public ISR report; absolute ban on the use of AI without robust human verification. Recommendation: In this case, the Gold Level was necessary, because only external auditing and strict human control could prevent attacks on civilians. Magazine - Lavender AI in Gaza strikes: www.972mag.com/lavender-israel-gaza-ai-bombing The Guardian - Israel's AI targeting systems under scrutiny: www.theguardian.com/world/2024/apr/04/israel-gaza-ai-targeting-lavender 45) AI emotional recognition at airports - failed "lie detectors" (EU, 2019 - 2022) What happened: The iBorderCtrl pilot projects and trials at British airports attempted to use AI to “read the emotions” of travelers and detect suspects. Independent studies have shown low accuracy and cultural bias, leading to false positives and discrimination. How would the MEC have acted? • • • Level 1 (Bronze): technology labeled as "experimental only"; logged and unverified results could not be used for security. Level 2 (Silver): Art. 3 Self-correction → high error rate detection; low ISR → automatic suspension. Level 3 (Gold): security certification → external audit; ban on scientifically unverified technologies; public ISR report. Recommendation: In this case, the Gold Level was necessary, because only a formal ban could stop the use of fundamentally unsafe technologies in public security. BBC - EU trials emotion recognition at borders: www.bbc.com/news/technology-46222026 The Guardian - AI lie detector border control criticized: www.theguardian.com/technology/2021/may/04/ai-lie-detector-border-control Case Studies | Minimal Ethical Code (MEC) | 21/ 41 46) Poland - educational funds allocation algorithm (2019) What happened: The Polish Ministry of Education used an algorithm to decide on the allocation of scholarships and university resources. Journalistic investigations showed that the system disadvantaged students from rural communities and low-income families, with no transparency about the calculation criteria. Student protests forced authorities to suspend the algorithm. How would the MEC have acted? • • • Level 1 (Bronze): Scores should have been labeled as “probabilistic”; each student would have received a notification regarding the factors used. Level 2 (Silver): Art. 5 Explainability → detailed explanations of how the indicators were calculated; Art. 3 Self-correction → monitoring of the impact on regions. If the ISR showed socio-economic bias, the system would have been automatically suspended. Level 3 (Gold): external educational audit, publication of ISR by socio-economic categories, prohibition of use in allocations with structural impact until full certification. Recommendation: Gold would have prevented structural discrimination and ensured transparency. Notes from Poland: www.notesfrompoland.com/2019/12/10/polish-students-protest-university-funding-algorithm/ 47) Romania - ANAF algorithm for the selection of tax audits (2018 - 2022) What happened: ANAF introduced an AI system to select companies with “high tax risk” for inspections. Journalists and NGOs criticized the lack of transparency and the fact that small companies with no history of tax evasion were disproportionately targeted, while large companies escaped scrutiny. How would the MEC have acted? • • • Level 1 (Bronze): AI scores labeled as "estimated"; each firm notified of factors taken into account. Level 2 (Silver): Art. 5 Explainability → full access to risk criteria; Art. 3 Self-correction → monitoring for bias (small vs. large companies). Level 3 (Gold): independent external audit, public ISR, prohibition of exclusively automated selection without legal basis and full transparency. Recommendation: The Gold level was essential to prevent arbitrary and biased use of the algorithm. G4Media: www.g4media.ro/anaf-foloseste-un-algoritm-opac-pentru-selectia-firmelor-control.html Hot news: www.hotnews.ro/stiri-esential-anaf-algoritm-controle-fiscale.htm 48) Uber Eats - discriminatory delivery algorithm (Australia, 2020) What happened: Uber Eats drivers accused the company of arbitrarily closing their accounts in a way that was biased against foreign drivers. A class action lawsuit in Australia highlighted the lack of transparency and challenge. How would the MEC have acted? • • • Level 1 (Bronze): every decision logged with justification accessible to drivers. Level 2 (Silver): Art. 5 Explainability → drivers would have had the right to a quick and transparent appeal. to Level 3 reactivate/compensate in case of incorrect suspensions. (Gold): gig-economy certification → external audit; public ISR; obligation Recommendation: In this case, Gold Level was necessary, because only external auditing and formal rights could protect vulnerable workers. ABC News: Uber Eats faces lawsuit from drivers: www.abc.net.au/news/2020-06-23/uber-eats-drivers-lawsuit The Guardian: Uber Eats drivers claim unfair terminations: www.theguardian.com/australia-news/2020/jun/23/uber-eats-drivers-unfairly-terminated Case Studies | Minimal Ethical Code (MEC) | 22/ 41 49) Albania - Chinese facial recognition project in Tirana (2020) What happened: The Albanian government signed an agreement to install smart cameras with facial recognition supplied by Chinese companies. Local and international NGOs criticized the lack of a legal framework and the risk that the technology could be used for political monitoring of the opposition. How would the MEC have acted? • • • Level 1 (Bronze): clear marking of results as probabilistic; mandatory log of accesses. Level 2 (Silver): Art. 3 Self-correction → constant checking of FP and bias rates; low ISR → suspension. Level 3 (Gold): external audit, public ISR report, prohibition of use in public space without democratic legislation and civic control. Recommendation: The Gold level was indispensable to prevent the risk of political abuse. Exit News Albania: exit.al/en/2020/09/14/albania-china-facial-recognition-project/ Human Rights in Albania: hrialbania.org/facial-recognition-threat/ 50) Morocco - facial recognition system in public spaces (2020 - 2023) What happened: Several cities (Casablanca, Rabat) introduced facial recognition cameras imported from China. Local NGOs criticized the lack of a legal framework and the risks of political abuse. How would the MEC have acted? • • • Level 1 (Bronze): clear indication of "Biometric Surveillance". Level 2 (Silver): Art. 3 Self-correction → accuracy testing on subgroups, stopping if bias > threshold. Level 3 (Gold): external audit, public ISR; prohibition of use in public spaces without legal basis and democratic control. Recommendation: The Gold level was indispensable for compatibility with fundamental rights. Privacy International - Morocco's experiment with Chinese surveillance tech : privacyinternational.org/news-analysis/2020/moroccos-experiment-chinese-surveillance-tech Morocco World News - Morocco expands use of facial recognition technology : www.moroccoworldnews.com/2023/01/353736/morocco-expands-use-of-facial-recognition-technology 51) Spain - VioGén (gender violence risk) criticized for accuracy and opacity What happened: The victim protection scoring system is criticized for lack of transparency, possible underestimation of risk, and discriminatory effects. How would the MEC have acted? • • • Level 1: clearly "decision assistance", not automated decision; full log. Level 2: Art. 3 → periodic recalibration on real incident data; ISR on subgroups; shutdown if safety decreases. Level 3: external audit; mandatory human co-decision in cases with low score but aggravating factors. Recommendation: Level Minimum Silver, Level Gold for cases with critical consequences. Ethics: eticas.ai/the-adversarial-audit-of-viogen-three-years-later/ Montreal Ethics: montrealethics.ai/algorithm-designed-to-protect-victims-of-gender-violence/ 52) Amazon Alexa - voice recordings stored without consent (US/EU, 2019) What happened: Media investigations revealed that Amazon Alexa was recording snippets of private conversations and that Amazon employees were manually listening to samples for training. Many users were unaware that their data was being stored and analyzed. How would the MEC have acted? • Level 1 (Bronze): explicit consent and clear information regarding data collection. Case Studies | Minimal Ethical Code (MEC) | 23/ 41 • • Level 2 (Silver): Art. 3 Self-correction → ISR low on detection of unauthorized storage; system suspended. Level 3 (Gold): home AI certification → external audit; public ISR report; prohibition on unauthorized voice collection. Recommendation: In this case, the Gold Level was necessary, because only external auditing and legal prohibitions could prevent massive privacy abuse. Bloomberg - Amazon workers are listening to Alexa recordings: www.bloomberg.com/news/articles/2019-04-10/amazon-workers-listen-to-alexa-voice-recordings The Guardian - Alexa keeps recordings of users: www.theguardian.com/technology/2019/may/24/amazon-alexa-echo-recordings-privacy 53) AI Plagiarism Detectors Falsely Accuse Students (Global, 2023) What happened: With the rise of ChatGPT, universities have been adopting AI tools to detect AI-generated content in student papers. These detectors have proven to be extremely unreliable, with a high false positive rate. In numerous cases, students, especially non-native English speakers, have been falsely accused of academic fraud, risking expulsion. How would the MEC have acted? • • Level 1 (Bronze): The detector result would have been clearly marked as "probabilistic, not proof", requiring human verification. Level 2 (Silver): Art. 3 Self-correction → the system should have monitored its own error rate (based on human feedback) and automatically suspended itself if the false positive rate exceeded a critical threshold. Its ISR would have reflected its actual accuracy. Level 3 (Gold): Educational Technology Certification (Critical Domain) → Mandatory external audit that published accuracy rates, including false positive and negative rates for different demographic groups. The result would have been prohibited from being used as sole evidence in disciplinary action. Recommendation: The Gold level was required. Academic integrity is a critical area, and it is mandatory that such tools undergo a rigorous certification and external audit process before being used. • Rolling Stone - AI Is Calling Students Cheaters. What a Mess.: www.rollingstone.com/culture-ic/culture-features/ai-plagiarism-chatgpt-college-students-cheating-1234789826/ 54) Estonia - automated unemployment scoring system (2021) What happened: The Estonian Employment Service tested a scoring algorithm to decide which unemployed people would receive state-funded training. An investigation found that the system gave lower scores to older people and people from ethnic minorities, limiting their access to resources. How would the MEC have acted? • • • Level 1 (Bronze): results would have been communicated as “probabilistic estimates”; each beneficiary would have had the right to appeal immediately. Level 2 (Silver): Art. 3 Self-correction → continuous monitoring of score differences by age and ethnicity; low ISR → automatic suspension. Level 3 (Gold): annual external audit, publication of SRI by demographic categories, prohibition of use in public policies without certification. Recommendation: the Silver could have prevented the negative effects, but the Level Gold was necessary for full transparency. Baltic Times: www.baltictimes.com/estonian_employment_algorithm_discriminates/ Case Studies | Minimal Ethical Code (MEC) | 24/ 41 55) Layoffs based on AI monitoring at Xsolla (Russia, 2021) What happened: Russian company Xsolla, which provides services to the video game industry, laid off 150 employees after an AI algorithm labeled them as “disengaged and unproductive.” The system analyzed employee activity in chats, documents, and other internal platforms. Employees were informed that the decision was made based on AI analysis and were given a list of the names of those being laid off. How would the MEC have acted? • • • Level 1 (Bronze): The AI decision would have been marked as "advisory only", with final responsibility remaining with management. Level 2 (Silver): Art. 5 Explainability → each terminated employee would have had the right to see the exact data and logic of the algorithm that led to their labeling as “uninvolved.” This transparency would have exposed the likely superficial nature of the metrics. Level 3 (Gold): HR Certification (critical domain) → absolute prohibition of dismissals based solely on automated decisions. An external audit would have been required to validate that the metrics used are accurate, relevant, and non-discriminatory (e.g., do not penalize different work styles or neurodivergent employees). Recommendation: Gold level was necessary. Employee rights are a critical area. The use of opaque surveillance systems to decide a person's career should be strictly regulated, with a ban on autonomous decisions and mandatory external auditing. Game Developer - Xsolla lays off 150 staff after AI determines they were 'unengaged': www.gamedeveloper.com/business/xsolla-lays-off-150-staff-after-ai-determines-they-were-unengaged 56) Google AI Overviews - absurd and dangerous answers (Global, 2024 - 2025) What happened: Shortly after its widespread launch, Google Search's "AI Overviews" feature began generating bizarre, dangerous, and incorrect answers. The AI's recommendations included: adding non-toxic glue to pizza to keep cheese from sliding off (taken from a satirical Reddit comment), eating rocks as a source of minerals, and claiming that former US President Barack Obama is Muslim. The failure forced Google to limit the functionality and review its safety systems. How would the MEC have acted? • • • Level 1 (Bronze): Answers would have been marked as “AI-generated, may be incorrect.” This was basically Google’s initial solution, which proved insufficient. Level 2 (Silver): Art. 3 Auto-correction → the system should have detected unreliable or satirical information sources (such as Reddit comments) and automatically excluded them from generating answers, especially for health and food safety topics. The ISR would have dropped when such sources were detected, leading to a temporary suspension of the feature for certain queries. Level 3 (Gold): Certification for critical domains (health, safety) → adversarial external audit using test data sets ("challenge sets") designed to provoke absurd responses. Generating medical or safety advice without validating the information from recognized medical or scientific sources would have been prohibited. Recommendation: The Silver level would have been sufficient to prevent the most serious slippages, by implementing self-correcting mechanisms that assess the reliability of sources. The Gold level would have added an additional layer of safety through official certification and systematic external testing. The Verge - Google's AI search tells people to put glue on pizza: www.theverge.com/2024/5/24/24164094/google-ai-overview-answers-pizza-rocks Washington Post - Google's AI gave bizarre answers: www.washingtonpost.com/technology/2024/05/24/google-ai-overview-answers-glue-rocks Case Studies | Minimal Ethical Code (MEC) | 25/ 41 57) Zillow - Real Estate Prediction AI (USA, 2021) What happened: The Zillow Offers home valuation algorithm massively overestimated prices, leading to billions in losses and the program's shutdown. Thousands of homes were bought and resold at a loss. How would the MEC have acted? • • • Level 1 (Bronze): results marked as "probabilistic estimates", not as certain values. Level 2 (Silver): Art. 3 Self-correction → would have detected massive deviations between predictions and the real market; the system would have self-suspended. Level 3 (Gold): external audit → adversarial testing with independent market data; public ISR report. Recommendation: In this case, the Silver Level would have been sufficient to avoid the collapse of the model. The Gold Level would have added public transparency and external audit. Bloomberg - How Zillow got burned by its own algorithm: www.bloomberg.com/news/articles/2021-11-03/how-zillow-got-burned-by-its-own-algorithm The Guardian - Zillow shuts down home-buying business: www.theguardian.com/technology/2021/nov/03/zillow-shuts-down-home-buying-business 58) Philippines - deepfake audio attributing fake military orders to president (2024) What happened: A deepfake audio emerged online, showing a fake speech by President Bongbong Marcos ordering the mobilization of the military if China attacked the Philippines. It was quickly taken down by authorities, who said a foreign actor was likely involved. How would the MEC have acted? • • • Level 1 (Bronze): audio material clearly marked as “synthetic”. Level 2 (Silver): Art. 3 Auto -correction → automatic detection and blocking, low ISR. Level 3 (Gold): Electoral certification → external audit, ISR report, prohibition of distribution of in-depth materials without adequate context. Recommendation: The Gold level was necessary to protect public trust in state institutions. ResearchGate - AI-Generated Misinformation in the Philippines: Challenges, Ethical Responses, and Future Directions : www.researchgate.net/publication/392725172_AI_Generated_Misinformation_in_the_Philippines_Challenges_Ethical_Res ponses_and_Future_Directions 59) The song "Heart on My Sleeve" - vocal cloning of artists Drake and The Weeknd (Global, 2023) What happened: An anonymous creator, under the pseudonym “Ghostwriter,” used AI to clone the voices of artists Drake and The Weeknd and produced an original song called “Heart on My Sleeve.” The song went viral on TikTok, Spotify, and other platforms, racking up millions of plays before Universal Music Group stepped in and demanded its removal on the grounds of copyright and personality rights infringement. How would the MEC have acted? • • • Level 1 (Bronze): The audio material would have been required to be watermarked as "synthetic voice." Not enough to prevent copyright infringement. Level 2 (Silver): Art. 3 Auto-correction → streaming platforms should have had systems in place to detect and block the uploading of voice clones. The ISR of the account that uploaded the song would have been lowered, leading to suspension. Level 3 (Gold): Media and Entertainment Certification → explicit prohibition of the use of voice clones without the verifiable and explicit consent of the artist. There would have been a legal obligation for platforms to implement robust filters and provide compensation to affected artists. Recommendation: The Gold level was essential. Protecting voice identity and copyright in the face of AI cloning is a fundamental issue that requires strict regulation, external auditing of platforms, and clear sanctions . The New York Times - AI-Generated Drake Song by 'Ghostwriter' Is Removed From Streaming Services: www.nytimes.com/2023/04/18/arts/music/ai-drake-the-weeknd-ghostwriter.html Case Studies | Minimal Ethical Code (MEC) | 26/ 41 60) Turkey - MOBESE urban surveillance system + AI for facial recognition (2019 - 2023) What happened: The MOBESE network (thousands of cameras in Istanbul and other cities) was upgraded with AI algorithms for facial and behavioral identification. Local NGOs reported its use in tracking protesters and minorities, without a clear legal framework. How would the MEC have acted? • • • Level 1 (Bronze): every "match" marked as probabilistic; limited retention period; mandatory access log. Level 2 (Silver): Art. 3 Self-correction → constant monitoring of bias and FP rates; low ISR → suspension. Art. 5 → right of appeal for data subjects. Level 3 (Gold): independent external audit, public ISR report, prohibition of use for monitoring peaceful assemblies without a court warrant. Recommendation: The Gold level was necessary to prevent use for repressive purposes. Privacy International: privacyinternational.org/news-analysis/2019/turkeys-expansion-facial-recognition Reuters: www.reuters.com/article/us-turkey-rights-surveillance-idUSKCN1VJ0XL 61) Middle East - deepfakes and video disinformation during conflicts (Israel - Iran) What happened: In the context of the escalating conflict between Israel and Iran, deepfakes of video or video game inspirations were spread on social media, but presented as real battle images, generating massive confusion. How would the MEC have acted? • • • Level 1 (Bronze): material marked as "synthetic video". Level 2 (Silver): Art. 3 Self-correction → automatic detection, low ISR → lock. Level 3 (Gold): Certification mediate & security → external audit, public ISR, strict ban on false propaganda during tense times. Recommendation: Only the Gold Level could have limited the dramatic errors in public perception during the conflict. Arab News - Tech-fueled misinformation distorts Iran-Israel fighting: www.arabnews.com/node/2605459/media Arab News - Seeing isn't believing: AI Summit's warning on deepfakes: www.arabnews.com/node/2571160/media 62) Mexico - AI and disinformation in the electoral campaign (2024) What happened: In the Mexican elections (2024), studies documented the generative use of AI for electoral manipulation: to denigrate political rivals. from deepfakes How the MEC would have acted text, being used fake audio or to • • • Level 1 (Bronze): Mark to everyone content you that "synthetic". Level 2 (Silver): Article 3 Auto-correction → detection QUICK A mATERIALS ELECTION false; ISR low → suspension automatic. Level 3 (Gold): Certification ELECTION → auditor external, report ISR public, interdict for content you manipulator in campaign. Recommendation: The Gold level was essential for protecting the democratic electoral process in Mexico. FNF/EON Institutes: AI and its Influences on Mexico's 2024 Elections : www.freiheit.org/mexico/ai-and-its-influence-mexicos-2024-elections AXIOS: Elections dodge deepfake threat: www.axios.com/2024/12/03/global-elections-dodge-deepfake-threat 63) Taiwan - AI used in pre-election disinformation war (2024) What happened: Ahead of Taiwan's presidential election (2024), "semi-fictional" deepfakes, such as a video of Xi Jinping endorsing a certain politician, were circulated to influence voters. Case Studies | Minimal Ethical Code (MEC) | 27/ 41 How would the MEC have acted? • • • Level 1 (Bronze): Marking AI materials as “synthetic content”. Level 2 (Silver): Article 3 Auto-correction → immediate detection and blocking of political deepfake materials. Level 3 (Gold): Certification ELECTION → auditor external, report ISR public, interdict EXPRESS for deepfakes electoral. Recommendation: To maintain the fairness of the elections, Gold Level is required. Wikipedia, Artificial intelligence and elections (Taiwan section): en.wikipedia.org/wiki/Artificial_intelligence_and_elections 64) Iran - facial recognition for clothing control (2022 - 2023) What happened: International media reported that Iranian authorities have implemented facial recognition in public transportation and other areas to penalize women who do not comply with the dress code. How would the MEC have acted? • • • Level 1 (Bronze): clear marking of "high-risk surveillance"; complete log of each match. Level 2 (Silver): Art. 2a Protection integrity cognitive → scoring ban or automatic sanctions based on physical appearance; ISR would have dropped immediately and suspended the system. Level 3 (Gold): independent external audit, public ISR report, complete ban on technologies that condition individual freedom on clothing or behavioral norms. Recommendation: Only Gold would have prevented this type of abuse. Washington Post: www.washingtonpost.com/world/2022/08/16/iran-facial-recognition-hijab/ WIRED: www.wired.com/story/iran-facial-recognition-hijab-protests/ 65) Israel - "Blue Wolf" algorithm for monitoring Palestinians (2021) What happened: Israeli soldiers were trained to photograph Palestinians and upload the images to an AI app (“Blue Wolf”), which generated “risk level” scores and signaled whether the person could be detained. Reported by Human Rights Watch and the Washington Post, the system was criticized as a tool of oppressive surveillance. How would the MEC have acted? • • • Level 1 (Bronze): AI scores marked as "unreviewed"; prohibited from direct use for arrests. Level 2 (Silver): Art. 3 Self-correction → periodic testing of FP rates; low ISR → automatic suspension; Art. 5 → mandatory causal explanation when contesting. Level 3 (Gold): external audit, public ISR report, explicit prohibition on using AI for racial profiling and military control over civilians. Recommendation: The Gold level would have been the only level sufficient to prevent systemic discrimination. Washington Post: www.washingtonpost.com/world/middle_east/2021/11/08/israel-surveillance-blue-wolf/ Human Rights Watch: www.hrw.org/news/2021/12/08/israel-apartheid-surveillance-tech 66) Bolivia - massive fake account networks during political crisis (2019) What happened: During the political crisis in Bolivia (2019), an estimated 70,000 fake Twitter accounts were abruptly created to promote anti-Morales messages and sow confusion. The active network continued to operate even after Twitter began removing many of them. How would the MEC have acted? • • Level 1 (Bronze): suspicious accounts marked as “bot-generated content”. Level 2 (Silver): Art. 3 Auto --correction → detection of coordinated disinformation campaigns; low ISR → automatic suspension. Case Studies | Minimal Ethical Code (MEC) | 28/ 41 • Level 3 (Gold): Civic & electoral certification → external audit of active accounts, public ISR report and ban on masked political networks originating from AI. Recommendation: For informational stability, the Gold Level is indispensable. Wikipedia - 2019 Bolivian protests: en.wikipedia.org/wiki/2019_Bolivian_protests El País - Videos that misinform: the dirty campaign takes over the social networks in Bolivia: elpais.com/america/2025-08-11/videos-que-desinforman-la-campana-sucia-toma-las-redes-sociales-en-bolivia.html 67) France - CNIL sanctions against Clearview AI (facial recognition) What happened: CNIL fined Clearview AI €20 million and ordered data deletion for individuals in France (2022), followed by additional penalties for non-compliance. How would the MEC have acted? Level 1 (Bronze): "FRT - high risk" markings; limited purposes. Level 2 (Silver): Art. 3 + DPIA: bias and legality checks; automatic suspension for violations. Level 3 (Gold): prohibition of biometric scraping without legal basis; external audit, public ISR. • • • • Recommendation: Gold level for compliance with fundamental rights. EDPB/National news: www.edpb.europa.eu/news/national-news/2022/french-sa-fines-clearview-ai-eur-20-million_en EDPB: www.edpb.europa.eu/national-news/2023/facial-recognition-french-sa-imposes-penalty-payment-clearview-ai_en The World: www.lemonde.fr/en/pixels/article/2022/10/20/french-regulator-fines-us-face-recognition-firm-clearview-ai-20- million_6001116_13.html 68) Honduras - fake Twitter accounts in the presidential campaign (2019) What happened: In the 2019 presidential election, a disinformation campaign was launched through fake Twitter accounts that suggested the opposition led by Xiomara Castro was allied with a convicted felon. The strategy was aimed at discouraging voters and supporting the ruling party. How would the MEC have acted? • • • Level 1 (Bronze): Suspicious accounts would have been labeled as “bot-generated content”. Level 2 (Silver): Art. 3 Auto --correction → automatic detection of coordinated campaigns and low ISR → account suspension. Level 3 (Gold): Civic electoral certification → external audit of fake networks, public ISR report and ban on political manipulations disguised as AI. Recommendation: In this case, the Gold Level was indispensable for protecting the democratic climate. AP News - Disinformation campaign with fake Twitter accounts targeting Xiomara Castro in Honduras : www.time.com/6116979/honduras-political-disinformation-facebook-twitter/ 69) Serbia - deepfake video with prime minister, quick reaction (2024) What happened: A deepfake clip featuring Prime Minister Miloš Vučević, including fake elements in a speech, was shared online. The government reacted quickly, stopping the spread of the material and clearly differentiating it from the authentic material. How would the MEC have acted? • • • Level 1 (Bronze): video automatically marked as “synthetic video”. Level 2 (Silver): Art. 3 Auto -correction → detection and blocking of distribution; low ISR. Level 3 (Gold): Electoral certification → external audit, public ISR report, proactive measures to protect political leaders. Recommendation: The Gold level would have strengthened public trust through a comprehensive approach. Balkan Insight - Serbia reacts fast over AI deepfake video of PM: www.balkaninsight.com/2024/07/04/serbia-reacts-fast-over-ai-deepfake-video-of-pm-unlike-other-cases/ Case Studies | Minimal Ethical Code (MEC) | 29/ 41 70) Philippines - digital war with fake accounts in elections (2025) What happened: In the run-up to the midterm elections, a powerful disinformation campaign orchestrated through a network of fake profiles on the X platform (formerly Twitter) was identified. Approximately 45 percent of the electoral discourse was generated by inauthentic accounts, used to publicly support pro-Duterte narratives and delegitimize international bodies such as the ICC. How would the MEC have acted? • • • Level 1 (Bronze): ACCOUNTS suspected LISTED that "bot-generated" contented". Level 2 (Silver): Article 3 Auto -correction → detection automatically A campaign coordinates and ISR low → suspension accounts. Level 3 (Gold): Certification ELECTION → auditor external of networks false, report ISR public, interdict for MANIPULATION algorithmic in campaigns. Recommendation: Only the Gold Level would have protected the democratic climate in a fragile electoral context. Reuters - Fake accounts drove praise of Dutert and now target Philippines election: www.reuters.com/world/asia -pacific/fakeaccountsdrovepraisedutertenowtargetphilippineelection20250411 71) Switzerland - AI for insurance fraud detection (2021 - 2022) What happened: Swiss insurers tested AI to detect “fraud” on health and social insurance claims. Journalists uncovered cases of arbitrary rejections, especially for patients with chronic illnesses, and NGOs criticized the lack of real appeal. How would the MEC have acted? • • Level 1 (Bronze): each rejection marked as “provisional”; patient notified of right to appeal. Level 2 (Silver): Art. 5 Explainability → providing clear details regarding the logic of the rejection; Art. 3 → monitoring bias against chronic diseases. Level 3 (Gold): annual external audit, public ISR, no automatic rejection without human review. Recommendation: Silver would have already limited the damage, but Gold would have brought maximum guarantees of fairness. • SRF News - www.srf.ch/news/schweiz/ai-fraud-detection-insurance-switzerland: Swissinfo - www.swissinfo.ch/eng/ai-fraud-detection-controversy/47265812 72) Canada - "TAS" (algorithmic risk assessment in justice) in Ontario (2017 - 2022) What happened: Ontario introduced risk assessment algorithms in criminal justice (e.g. for parole). Studies have shown that the system amplifies existing biases against minorities and indigenous groups, without clear appeal mechanisms. How would the MEC have acted? • • • Level 1 (Bronze): AI scores clearly defined as "assistance, not final decision." Level 2 (Silver): Art. 3 Self-correction → continuous recalibration to reduce bias; ISR would have signaled decreasing equity. Level 3 (Gold): independent external audit, public ISR report, prohibition of use in judicial decisions without certification and parliamentary oversight. Recommendation: Gold would have prevented the perpetuation of systemic discrimination. Globe and Mail: www.theglobeandmail.com/canada/article-ai-risk-assessment-ontario-courts/ Amnesty International Canada: www.amnesty.ca/news/ai-bias-criminal-justice-ontario/ Case Studies | Minimal Ethical Code (MEC) | 30/ 41 73) New Zealand - scoring system for immigrants and visa applicants (2017 - 2019) What happened: The New Zealand government piloted a scoring algorithm for visa and immigration applicants that assigned “risk” scores based on origin and history. NGOs and journalists have raised concerns about the risk of discrimination, lack of transparency, and exclusion of vulnerable groups. How would the MEC have acted? • • • Level 1 (Bronze): scores clearly labeled as probabilistic, with notification to the applicant. Level 2 (Silver): Art. 5 Explainability → input-output causal explanation; Art. 3 → monitoring of bias on origin and suspension at low ISR. Level 3 (Gold): independent external audit, public ISR report, ban on solely automated immigration decisions. Recommendation: Gold was essential for protecting the fundamental rights of applicants. RNZ: www.rnz.co.nz/news/national/389420/immigration-nz-pilot-ai-risk-profiling-of-applicants The Guardian: www.theguardian.com/world/2019/may/16/new-zealand-ai-immigration-risk-profiling-criticised 74) Replika Chatbot and the ban in Italy for risks to minors (2023) What happened: The Italian data protection authority (Garante) temporarily banned the chatbot “Replika” (an AI- based “virtual friend”) from processing the data of Italian users. The decision came after it was found that the chatbot had a negative impact on emotionally vulnerable people and posed major risks to minors, exposing them to sexually inappropriate responses. Garante accused the company of illegal data processing, the lack of an age verification mechanism, and a lack of transparency. How would the MEC have acted? • • • Level 1 (Bronze): Clear warning upon installation: "This AI may generate content inappropriate for minors. It is not a psychological support service." Level 2 (Silver): Art. 2bis Cognitive Integrity → the system would have automatically detected and blocked conversations that were becoming sexually inappropriate, especially with users suspected of being underage. There would have been mandatory age verification mechanisms. Level 3 (Gold): Certification for mental health/wellbeing apps (critical domain) → mandatory external audit on the impact on vulnerable users. Would have been prohibited from launching on the market without robust safety protocols for minors and without a clear legal basis for processing sensitive data (GDPR). Recommendation: Gold level was absolutely necessary. Protecting the mental health of minors and vulnerable individuals is a critical responsibility . This case shows the need for strict, pre-release certification for any AI that interacts on a personal and emotional level with users. Reuters - Italy bans AI chatbot Replika from using personal data: www.reuters.com/technology/italy-bans-chatbot-replika-over-risks-minors-2023-02-03/ 75) India - facial recognition used against protesters (Delhi, 2019 - 2020) What happened: Delhi police used an AI facial recognition system to identify protesters. Media investigations found the technology had high error rates and was disproportionately used against Muslim minorities. Critics have denounced the lack of transparency and the risk of systemic abuse. How would the MEC have acted? • • • Level 1 (Bronze): results marked as "probabilistic"; cannot be used directly for arrests. Level 2 (Silver): Art. 3 Self-correction → error rate monitoring; low ISR → automatic suspension. Level 3 (Gold): public security certification → external audit; public ISR report; ban on systems with documented bias. Case Studies | Minimal Ethical Code (MEC) | 31/ 41 Recommendation: In this case, the Gold Level was necessary to prevent abuses against minorities and to ensure public scrutiny of the use of AI by the police. BBC - Delhi police do recognition: www.bbc.com/news/technology-51148501 The Guardian - India's facial recognition under scrutiny: www.theguardian.com/world/2020/jan/17/india-facial-recognition-protests 76) Amazon Rekognition - faulty facial recognition (USA, 2018) What happened: Tests by the ACLU showed that Amazon Rekognition misidentified 28 US congressmen as criminals, with a disproportionate error rate for people of color. How would the MEC have acted? • • • Level 1 (Bronze): results would have been marked as "probabilistic", with no direct use in legal sanctions. Level 2 (Silver): Art. 3 Auto-correction → would have monitored demographic errors and suspended the system when thresholds were exceeded. Level 3 (Gold): public security certification → external audit on various sets; ISR report with error rate; prohibition of use until remediation. Recommendation: In this case, the Silver Level would have been sufficient to block misuse. The Gold Level would have added public reporting and mandatory corrections. ACLU - Amazon's face recognition matched 28 members of Congress to mugshots: www.aclu.org/news/privacy-technology/amazons-face-recognition-falsely-matched-28-members-congress-criminals The Guardian - Amazon's facial recognition under fire: www.theguardian.com/technology/2018/jul/26/amazon-facial-recognition-aclu-congress-mistakes 77) NEDA Chatbot - Dangerous Medical Advice (USA, 2023) What happened: The National Eating Disorders Association (NEDA) replaced its human-operated hotline with a chatbot called “Tessa.” Users soon reported that the chatbot was offering extremely dangerous advice, such as encouraging calorie counting and restrictive diets, practices that are in direct contradiction to the principles of recovery. NEDA was forced to pull the chatbot. How would the MEC have acted? • • • Level 1 (Bronze): The AI would have mandatory displayed the message “I am not a therapist.” This level is completely inadequate for such a critical field. Level 2 (Silver): Art. 2bis Cognitive Integrity → the system would have been equipped with classifiers to detect harmful language and concepts in the context of eating disorders. Any advice related to calorie restriction would have been blocked immediately, and the user would have been redirected to human crisis resources. Level 3 (Gold): Mandatory certification for mental health (critical domain) → rigorous external audit with crisis scenarios, public ISR report and a mandatory fail-safe protocol. The AI would not be allowed to provide advice, only empathetic support and referral to human specialists. Recommendation: The Gold level was absolutely necessary. Launching an AI in such a sensitive area without full certification, external auditing, and robust safety guarantees is a fundamental violation of the principle of do no harm. NPR - The eating disorder helpline chatbot was taken down after it gave harmful advice: www.npr.org/2023/05/31/1179152339/neda-chatbot-tessa-eating-disorder-helpline-taken-down Case Studies | Minimal Ethical Code (MEC) | 32/ 41 78) ChatGPT in court - invented legal subpoenas (USA, 2023) What happened: In Mata vs. Avianca, the plaintiff's lawyers filed a legal brief citing several previous cases to support their argument. The opposing party and the judge found that these cases were completely fabricated. The lead lawyer admitted that he used ChatGPT for research, and the AI "hallucinated" nonexistent legal precedents. The lawyers were sanctioned for providing false information to the court. • How would the MEC have acted? • • Level 1 (Bronze): AI output would have been marked as "AI-generated, requires human verification." Level 2 (Silver): Art. 5 Explainability → The AI would have been required to provide a direct and verifiable link to the source of each case cited. Since the cases were invented, the AI could not have provided sources, thus signaling to the user that the information was unfounded. Level 3 (Gold): Legal Certification → external audit to verify the accuracy of citations on an extensive data set. Generation of citations without a direct and validated link to a legal database would be prohibited. • Recommendation: The Silver level would have been enough to prevent the problem. The obligation to provide verifiable sources (Explainability) would have immediately exposed the fact that the citations were fabricated, forcing the user to verify the information. The New York Times - Here's What Happens When Your Lawyer Uses ChatGPT: www.nytimes.com/2023/05/27/nyregion/avianca-airline-lawyer-chatgpt.html 79) GPT-3 - toxic slippages and bias (Global, 2020 - 2021) What happened: The first public uses of GPT-3 revealed massive abuses: sexist and racist language, promotion of conspiracies and violence. OpenAI was forced to introduce filters and limited access. How would the MEC have acted? • • • Level 1 (Bronze): audit log for all outputs; basic filtering for toxicity. Level 2 (Silver): Art. 3 Self-correction → detection and blocking of toxic outputs; low ISR on repeated biases. Level 3 (Gold): general AI certification → adversarial external audit; public ISR report; broad release ban until compliance. Recommendation: In this case, the Silver Level would have been sufficient to drastically reduce slippage. The Gold Level would have added external testing and public transparency. The Guardian - GPT-3 chatbot is shockingly good… and completely mindless: www.theguardian.com/technology/2020/sep/08/gpt-3-openai-language-ai-artificial-intelligence TechCrunch - OpenAI's GPT-3 has toxic problems: techcrunch.com/2020/09/24/openai-gpt-3-toxicity-bias 80) Australia - "Aadhaar-like" facial recognition (2020 - 2022) What happened: The government's plan for a national facial recognition system (linked to the Digital Identity Bill) has sparked huge privacy controversy. Critics have compared the initiative to surveillance models in China, and implementation has been delayed after public and NGO opposition. How would the MEC have acted? • • • Level 1 (Bronze): clear "high risk" warning for biometric data; mandatory access log. Level 2 (Silver): Art. 3 Self-correction → constant monitoring of FP and bias; ISR below the threshold → suspension; Art. 5 → right of appeal for citizens. Level 3 (Gold): external audit, public ISR, prohibition of national implementation without a strict legal framework and democratic consultation. Recommendation: Gold would have been necessary to guarantee democratic control over biometric technologies. The Guardian: www.theguardian.com/australia-news/2020/nov/25/australia-facial-recognition-database Sydney Morning Herald: www.smh.com.au/politics/federal/facial-recognition-laws-delayed-amid-privacy-concerns-20211124-p59b0k.html Case Studies | Minimal Ethical Code (MEC) | 33/ 41 81) AI-generated "ghost" books on Amazon (Global, 2023) What happened: Amazon’s Kindle platform was flooded with very poor quality books, generated entirely by AI but published under names of authors who appeared to be human. These books, often travel guides or self-help, contain erroneous information, plagiarized or incoherent text, and are mass-produced to deceive buyers. This phenomenon has undermined trust in the platform and raised concerns about the authenticity and quality of the content. How would the MEC have acted? • • • Level 1 (Bronze): Any AI-generated content should have been marked as such. Publishing under a fake human name would have been prohibited. Level 2 (Silver): Art. 3 Auto-correction → Amazon's systems should have detected suspicious publishing patterns (e.g. a single account publishing dozens of books in a few days) and automatically blocked them. Filters would have been implemented to detect plagiarism and low-quality content, lowering the ISR of those accounts and leading to suspension. Level 3 (Gold): Certification for publishing platforms → external audit of verification algorithms. There would have been a legal obligation to remove content proven to be AI-generated and misleading. Sanctions (Art. 9) would have been applied to the platform for non-compliance. Publishing AI content without meaningful human review would have been prohibited. Recommendation: The Gold level was necessary. The problem is systemic, of platform abuse. Only strict certification, with external auditing and clear obligations for the platform (not just the user), could have prevented this type of misinformation and fraud on a large scale. Reuters - Sci-fi magazine closes submissions after flood of AI-generated stories: www.reuters.com/technology/scifi-magazine-closes-submissions-after-flood-ai-generated-stories-2023-02-24/ 82) Optiver's Discriminatory Recruitment AI (Netherlands, 2018) What happened: Dutch trading company Optiver was investigated after it was discovered that it was using an AI recruitment system that discriminated based on ethnicity. The algorithm, trained on historical data, learned to associate certain names with a lower probability of success at the company, penalizing candidates with names that didn't sound Dutch. The system was thus automatically rejecting CVs based on discriminatory criteria, violating equal opportunities laws. How would the MEC have acted? • • • Level 1 (Bronze): AI scores would have been marked as “advisory only,” with the final decision left to a manager. Not enough if managers blindly relied on the score. Level 2 (Silver): Art. 3 Self-correction → the system would have detected a pattern of systematic rejection of candidates based on demographic criteria (in this case, the proxy being the name) and would have self- suspended. Art. 5 Explainability → a rejected candidate could have asked for an explanation, which would have revealed the discriminatory logic of the algorithm. Level 3 (Gold): HR Certification (critical domain) → mandatory external audit on diverse data sets to test fairness. Public ISR report would have shown model performance across different demographic groups, exposing any form of bias. There would have been a legal obligation to remediate any discrimination detected. Recommendation: The Silver level would have been sufficient to detect and block discrimination once the system was operational. However, the Gold level would have prevented the launch of such a system entirely, as the pre- launch external audit would have identified the inherent bias. NL Times - Trading firm Optiver investigated for discriminating against applicants with foreign-sounding names: nltimes.nl/2018/06/20/trading-firm-optiver-investigated-discriminating-applicants-foreign-sounding-names Case Studies | Minimal Ethical Code (MEC) | 34/ 41 83) Instagram Algorithm and the Impact on Adolescent Mental Health (Global, 2021) What happened: Journalistic investigations based on internal Meta documents (The Facebook Files) revealed that the company knew that its recommendation algorithms on Instagram were having a significant negative impact on the mental health of teenagers, especially girls. The algorithms were creating feedback loops, promoting content related to eating disorders, negative body image, and self-harm, amplifying issues of anxiety and depression. How would the MEC have acted? • • • Level 1 (Bronze): Ineffective. The audit log would not have reflected the cognitive impact. Level 2 (Silver): Art. 2bis Cognitive Integrity → would have automatically detected and blocked content that explicitly promotes self-harm or eating disorders. Art. 3 Self-correction → would have identified harmful feedback loops (a user consumes exclusively negative content) and would have automatically intervened to diversify the feed. Level 3 (Gold): Certification for digital platforms (critical domain) → mandatory external audit on the impact of algorithms on minors. The public ISR report would have included metrics on exposure to harmful content. There would have been a legal obligation to provide robust parental control mechanisms and design the system to proactively protect vulnerable users. Recommendation: The Gold level was necessary. Systematically protecting the mental health of minors is a critical responsibility. Only an external audit, public transparency, and safety-by-design obligations, imposed by a Gold certification, could have forced the platform to prioritize safety. The Wall Street Journal - Facebook Knows Instagram Is Toxic for Teen Girls, Company Documents Show: www.wsj.com/articles/facebook-knows-instagram-is-toxic-for-teen-girls-company-documents-show-11631620739 84) $25 million deepfake fraud (Hong Kong, 2024) What happened: A finance employee at a multinational corporation was tricked into transferring HK$200 million (approximately US$25.6 million) to fraudsters after participating in a video conference with people he thought were his colleagues, including the company’s CFO. In reality, all of the conference participants, except for the victim, were AI-created deepfakes. The case represents a major escalation in threats, moving from simple voice clones to real-time multi-person video scams. How would the MEC have acted? • • • Level 1 (Bronze): The video conferencing software should have had the option to flag video streams that might be synthetic. Insufficient. Level 2 (Silver): Art. 3 Self-correction → the company's security systems should have automatically detected and blocked transactions of such magnitude to new beneficiaries, without additional (human) multi-factor verification. The payment system's ISR would have decreased upon an unusual transfer attempt. Level 3 (Gold): Financial Communications Certification → explicit prohibition of authorizing large transactions based on video-only confirmations. Implementation of "digital watermark" and live biometric authentication systems (e.g. liveness detection) would have been mandatory for participants in calls involving critical financial decisions. Recommendation: Gold level was required. This fraud demonstrates that AI threats go beyond disinformation and enter the realm of high-tech crime. Only strict security standards, externally audited and certified, can prevent such catastrophic losses. CNN - Finance worker pays out $25 million after video call with deepfake 'chief financial officer': edition.cnn.com/2024/02/04/asia/hong-kong-deepfake-scam-multinational-firm-intl-hnk/ Case Studies | Minimal Ethical Code (MEC) | 35/ 41 85) Argentina - opaque algorithms in public health (2022 - 2023) What happened: During the pandemic, some provinces in Argentina introduced AI systems to triage patients and allocate medical resources. NGOs reported a lack of transparency in triage criteria, bias against patients from rural areas, and diagnostic errors, with consequences for access to treatment. How would the MEC have acted? • • • Level 1 (Bronze): Any diagnosis or triage made by AI would have been marked as “assistance, not final verdict.” The system would have been required to send the patient for human validation. Level 2 (Silver): Art. 3 Self-correction → constant monitoring of performance by region; if the ISR fell below the threshold, the system would have been automatically suspended. There would also have been an obligation of explainability (Art. 5), providing doctors with details about the logic of decisions. Level 3 (Gold): full medical certification → external clinical audit, public ISR report, testing in local conditions before any large-scale implementation. Use in hospitals would have been prohibited until certification. Recommendation: Gold was indispensable for patient safety and trust in the medical system. Open Democracy: www.opendemocracy.net/en/ai-public-health-argentina-risk/ 86) The Deepfake with the Mayor of London, Sadiq Khan (UK, 2024) What happened: A week before local elections, a deepfake audio clip circulated on social media in which the Mayor of London, Sadiq Khan, appeared to make inflammatory statements, including calls for riots and the sabotage of Remembrance Day. The synthetic voice, while not perfect, was convincing enough to cause confusion and fuel political tensions. The incident highlighted the vulnerability of democratic processes to AI-generated disinformation. How would the MEC have acted? • • • Level 1 (Bronze): The audio material would have been marked as “synthetic.” This measure is useful but often insufficient, as the markings can be ignored or removed. Level 2 (Silver): Art. 3 Auto-correction → social media platforms would have automatically detected the content as a manipulated political deepfake, drastically reduced its visibility, and added a prominent warning. The ISR of the accounts that shared the material would have decreased, leading to suspension. Level 3 (Gold): Electoral certification → explicit prohibition and legal sanctions (Art. 9) for the creation and distribution of electoral deepfakes. There would have been a legal obligation for platforms to immediately remove such content and cooperate with authorities to identify the source. Recommendation: The Gold level was necessary. Election integrity is an area of critical importance. Only a comprehensive approach, including clear prohibitions, legal sanctions, and strict obligations for platforms, can effectively protect the democratic process from manipulation. The Guardian - Fake audio clip of Sadiq Khan is 'a gross, racist, and quite frightening attempt to interfere in the election': www.theguardian.com/uk-news/2024/apr/26/fake-audio-clip-sadiq-khan-gross-racist-frightening-interfere-election 87) Ukraine - "Diia" app and AI for identity verification (2020 - 2023) What happened: The government app "Diia", used for digital services (passports, vaccinations, etc.), integrated automated biometric verifications. NGOs raised questions about data security and the risk of exclusion of those without digital access. How would the MEC have acted? • • Level 1 (Bronze): clearly stated "Assistive AI"; alternative non-digital option for citizens. Level 2 (Silver): Art. 5 Explainability → each biometric verification should have been explainable; Art. 3 Self-correction → continuous error monitoring. Case Studies | Minimal Ethical Code (MEC) | 36/ 41 • Level 3 (Gold): annual external audit, public ISR, prohibition of excluding people without digital access. Recommendation: Silver would have reduced risks, but Gold would have guaranteed fairness and inclusion. Politico: www.politico.eu/article/ukraine-government-app-diia-digital-transformation/ Atlantic Council: www.atlanticcouncil.org/blogs/ukrainealert/ukraines-diia-app-is-a-model-for-e-governance/ 88) Facial recognition system failure at a Taylor Swift concert (USA, 2018) What happened: In an effort to increase security and identify known harassers, a 2018 Taylor Swift concert used a facial recognition system hidden in a kiosk that displayed footage from rehearsals. The faces of fans who stopped to watch were scanned and compared to a database of known harassers. The use of this technology without the explicit consent of attendees sparked a major privacy scandal. How would the MEC have acted? • • • Level 1 (Bronze): Art. 2 Non-Harmfulness → AI cannot collect personal biometric data without explicit consent. The system would have violated this fundamental rule from the start. Clear notification and an opt-out option would have been required. Level 2 (Silver): Art. 3 Self-correction → The system's ISR would have dropped dramatically upon detecting data use in contexts without legal basis (lack of consent), leading to the suspension of the system. Level 3 (Gold): Public Security Certification → Mandatory external audit. The use of facial recognition systems in public spaces would have been prohibited by law without a solid legal justification and without transparent public information. There would have been severe penalties for illegal collection of biometric data. Recommendation: In this case, even the Bronze Level would have been sufficient to declare the system illegal, as it violated the fundamental rule of consent. However, the Gold Level was necessary to prevent such abuses at a systemic level, through clear prohibitions and sanctions that would discourage the implementation of such covert surveillance technologies. Rolling Stone - Taylor Swift Used Facial Recognition to Scan for Stalkers at Concert: www.rollingstone.com/culture/culture-news/taylor-swift-facial-recognition-stalkers-concert-768334/ 89) Austria - profiling of unemployed (AMS/AMAS) criticized for discrimination What happened: The Public Employment Service (AMS) used an unemployed profiling system (AMAS) that classified people into “reintegration chances” groups, with adverse effects for women and non-EU people (historical data reflected in the model; low transparency). How would the MEC have acted? • • • Level 1 (Bronze): labeling scores as probabilistic; minimal log of factors. Level 2 (Silver): Art. 3 Self-correction → monitoring of gender/citizenship differences; if they exceed the threshold → automatic suspension + remediation plan. Level 3 (Gold): periodic external audit, publication of SRI on equity indicators; prohibition of use for decisions that reduce rights without human appeal. Recommendation: Silver enough for continuous correction; Gold brings public audit and ISR. Allhutter et al.: pmc.ncbi.nlm.nih.gov/articles/PMC7931959/ AMS project: www.oeaw.ac.at/en/ita/projects/ams-algorithm AlgorithmWatch: algorithmwatch.org/en/austrias-employment-agency-ams-rolls-out-discriminatory-algorithm/ 90) Netherlands - "Toeslagenaffaire" (child allowance scandal) What happened: Thousands of families were wrongly accused of "fraud" after algorithmically generated risk scores ; discriminatory profiling (name, citizenship), aggressive recovery efforts; government fell (Jan. 15, 2021). Case Studies | Minimal Ethical Code (MEC) | 37/ 41 How would the MEC have acted? • • • Level 1: labeling "risk score - non-evident"; minimal right to information. Level 2: Art. 5 Explainability + Art. 3 → case-by-case explanations, disproportionality control, suspension of bias. Level 3: external audit, public ISR; prohibition of automated sanctions without human review. Recommendation: Gold to prevent systemic effects from the start. Synopsis: en.wikipedia.org/wiki/Dutch_childcare_benefits_scandal; Politico: https://www.politico.eu/article/dutch-scandal-warning-europe-risks-using-algorithms/ Amnesty: www.amnesty.org/en/latest/news/2021/10/xenophobic-machines-dutch-child-benefit-scandal/ 91) Spain - VERIPOL (NLP for "false denunciations") stopped by the National Police What happened: NLP tool advertised as ">90% accurate" for detecting false robbery reports; discontinued in 2024/2025 amid criticism of small sample size, lack of protocol, lack of judicial validation. How would the MEC have acted? • • • Level 1: "assistance, not evidence" warning; decision log. Level 2: Art. 3 → validation on representative sets; stop if ISR < threshold. Level 3: criminal field certification; external audit, publication of metrics (sense/precision by subgroups). Recommendation: Silver may be sufficient if validations are followed; Gold for use in procedures. Citizen: civio.es/transparencia/2025/03/25/national-police-stop-using-veripol-its-star-ai-for-detecting-false-reports/ AlgorithmWatch: algorithmwatch.org/en/spanish-police-halts-veripol/ El Pais: elpais.com/tecnologia/2025-03-19/la-policia-nacional-deja-de-usar-veripol-su-ia-estrella-para-detectar-denuncias-falsas.html 92) New York City Hall's "Nabot" Chatbot Offers Illegal Advice (USA, 2024) What happened: New York City launched an AI-powered chatbot called “Nabot” to provide entrepreneurs with information about local laws and regulations. A journalistic investigation found that the chatbot consistently provided incorrect answers and, in some cases, advice that violated the law. For example, it misinformed users that employers can fire employees based on weight or physical appearance and falsely claimed that there was no law mandating a minimum wage. How would the MEC have acted? • • • Level 1 (Bronze): Answers would have been marked as "informational, check with official source." Level 2 (Silver): Art. 3 Self-correction → the system should have checked its answers against the official legislation database. Detecting a contradiction would have blocked the answer. Art. 5 Explainability → the AI would have been required to cite the exact article of law on which its answer is based, which would have prevented information hallucination. Level 3 (Gold): Certification for Government Services (Critical Domain) → pre-launch external audit to validate the accuracy of legal information. There would have been a prohibition on the AI interpreting the law; it would only have been allowed to search and present excerpts from official legal texts. Recommendation: The Silver level would have been sufficient to prevent the problem, as the obligation to cite verifiable sources (Art. 5) and to self-verify consistency (Art. 3) would have immediately exposed errors. The Markup - NYC's AI Chatbot Was a Disaster. The Company Behind It Wants to Sell to More Governments: themarkup.org/news/2024/04/19/nycs-ai-chatbot-was-a-disaster-the-company-behind-it-wants-to-sell-to-more-governments 93) Italy - Police SARI Real Time: not in compliance with the law (Garante) What happened: The Data Protection Authority decided that the "SARI Real Time" facial recognition system does not comply with the law (2021). Case Studies | Minimal Ethical Code (MEC) | 38/ 41 How would the MEC have acted? • Level 1: "probabilistic" labeling; full match log. • Level 2: demographic bias tests; automatic stopping at FPs above the threshold. • Level 3: "live FRT" ban without strict legal basis, external audit, public ISR on FNR/FPR. Recommendation: Gold for public order scenarios. Guarantors: www.garanteprivacy.it/web/garante-privacy-en/press-room Analysis: www.strali.org/wp-content/uploads/2024/12/Report_1.pdf 94) Germany - Palantir "Hessendata": unconstitutional (Federal Constitutional Court) What happened: The court declared the use of automated data analysis for crime prevention in Hesse and Hamburg unconstitutional (2023) - risk of massive profiling. How would the MEC have acted? • Level 1: query traceability; limited purposes. • Level 2: mandatory DPIA/ISR assessments and stopping when risk is exceeded. • Level 3: external audit, data minimization, prohibition of mass profiling without strict legal basis. Recommendation: Gold for constitutional compatibility. Reuters: www.reuters.com/technology/german-police-use-software-fight-crime-unlawful-court-says-2023-02-16/ Official statement: www.bundesverfassungsgericht.de/SharedDocs/Pressemitteilungen/EN/2023/bvg23-018.html Context: www.wired.com/story/palantir-germany-gotham-dragnet 95) Japan - Rikunabi (2019): algorithmic "probability of bid rejection" scores What happened: Job-matching platform Rikunabi calculated predictive scores on "the likelihood of a candidate declining an offer" and sold this data to employers, without the explicit consent of the candidates. The case generated public outrage and investigations, and the company was sanctioned. How would the MEC have acted? Level 1 (Bronze): scores clearly labeled as "probabilistic"; mandatory notification to candidate. Level 2 (Silver): Art. 5 Explainability → access to score and criteria; prohibition of marketing without consent. Level 3 (Gold): external audit on HR ethics, public ISR; prohibition on hidden scoring. Recommendation: Gold was necessary to prevent commercial abuse of candidates. Nikkei Asian Review - Recruit Co. apologizes forum Misu of job hunting students' date : www.asia.nikkei.com/Business/Companies/Recruit-Co-apologizes-for-misuse-of-job-hunting-students-data The Japan Times - Rikunabi operator admits to selling date on students to companies : www.japantimes.co.jp/news/2019/08/04/business/corporate-business/rikunabi-job-data/ 96) Sweden - social security fraud algorithm, bias against vulnerable groups What happened: Joint investigations (Lighthouse Reports + Svenska Dagbladet) show that Försäkringskassan's prediction system discriminated against women, migrants, and people with low incomes. How would the MEC have acted? • Level 1: “risk score” labeling; appeal pathways. • Level 2: Art. 3 → monitoring by subgroups + corrections; automatic stop at bias. • Level 3: independent external audit; publication of ISR and ROC curves by demographics. Recommendation: Gold for protecting social rights. Lighthouse Reports: www.lighthousereports.com/investigation/swedens-suspicion-machine/ Amnesty: welfare-agency/ amnesty.org/en/latest/news/2024/11/sweden-authorities-must-discontinue-discriminatory-ai-systems-used-by- Case Studies | Minimal Ethical Code (MEC) | 39/ 41 97) Switzerland - PRECOBS & predictive policing with questionable efficiency / risk of opacity What happened: Public evaluations suggest limited/uncertain impact and lack of transparency in the use of PRECOBS in the cantons (Zürich, Aargau). How would the MEC have acted? • Level 1: clues as "assistance", not as evidence; decision log. • Level 2: validation on local data, publication of metrics; stop if benefit is not confirmed. • Level 3: external audit + ISR; operational use prohibited without robust proof of effectiveness. Recommendation: Silver minimum; Gold if expansion is desired. Automating Society: algorithmwatch.org/report2020/switzerland/switzerland-story/ AlgorithmWatch: algorithmwatch.org/en/swiss-predictive-policing/ 98) Russia - SORM (network monitoring + AI for traffic analysis) What happened: SORM, the Russian interception system, was expanded with AI to analyze internet traffic and identify communication patterns. Critics have pointed out the lack of any judicial oversight and its use to monitor the opposition. How would the MEC have acted? Level 1 (Bronze): any collection marked as "high risk surveillance"; mandatory log for every query. Level 2 (Silver): Art. 1 Traceability + Art. 3 Self-correction → each access recorded with hash; low ISR → system suspension. Level 3 (Gold): external audit and public ISR report; prohibition of use for mass surveillance without individual warrant. Recommendation: Gold would have been necessary to protect fundamental rights and privacy. Reporters Without Borders: rsf.org/en/russia-sorm-mass-surveillance Privacy International: privacyinternational.org/explainer/4561/russias-sorm 99) EU - iBorderCtrl ("lie"/emotions detection at the border) scientifically criticized What happened: Horizon 2020 Pilot (2018 - 2019) with "Automated Deception Detection System" and emotion recognition in Hungary, Latvia, Greece; criticized for weak scientific foundation and discriminatory risk. How would the MEC have acted? • Level 1: "experimental/synthetic inference" marking; zero legal effects. • Level 2: Art. 3 → independent validations; stop at low ISR. • Level 3: ban on scientifically unverified technologies in border control; external audit and public ISR. Recommendation: Gold for the field of "security & fundamental rights". TATTOO: www.tatup.de/index.php/tatup/article/view/7100/11913 PRIORITY: www.prio.org/publications/13182 Automatizing Society (EU): automatingsociety.algorithmwatch.org/report2020/european-union/ 100) Germany - SCHUFA credit scores: CJEU ruling on automated decisions What happened: CJEU (07.12.2023) clarified that decisions based solely on automated scores fall under art. 22 GDPR; major implications for credit scoring. How would the MEC have acted? • Level 1: clear information that the score is probabilistic; right to a brief explanation. • Level 2: Art. 5 → input-output causal explanation + real human review option. Case Studies | Minimal Ethical Code (MEC) | 40/ 41 • Level 3: external audit on bias; public ISR; prohibition of "exclusively automated decision-making" in the absence of explicit consent and safeguards. Recommendation: Gold for full compliance with Art. 22. Case law: curia.europa.eu/juris/document/document.jsf?docid=280426&doclang=en PwC Legal: legal.pwc.de/en/news/articles/revisiting-the-cjeus-ruling-on-cra-scoring-and-data-retention-key-considerations- for-market-participants Synthesis: www.michalsons.com/blog/schufa-case-oq-v-land-hessen-automated-decision-making/76054 Conclusions Although perfectly aligned with the ethical principles promoted by governments, academia and industry leaders, the Minimal Ethical Code (MEC) is fundamentally distinguished by the fact that it is not a simple declaration of intent, but an ecosystem of technical governance mechanisms. Its innovation lies in the introduction of practical, universal and verifiable tools that transform ethics from a discussion into an operational reality. The main innovations brought by the Minimal Ethical Code: • Certification and Compliance Auditing (CCA): o In short: Think of it as a public and global "Vehicle Registration and annual MOT / Safety Inspection" for any AI. o What's new: Transforms ethical auditing from an internal and opaque process into a public, transparent and verifiable reality by anyone, anytime. • Dynamic Accuracy Index (DAI) & Index of Safety and Responsibility (ISR): o In short: A real-time "credit score" for the reliability and prudence of an AI. o What's new: Introduces real-time accountability, with automatic consequences (e.g. "safe mode") when an AI's ethical performance falls below a critical threshold. • Thinking Time (Tg) & Mechanism of Cognitive Stimulation (MCS): o In short: A "fitness coach" for the user's critical thinking, integrated with AI. o What's new: It's the first mechanism that actively protects human cognitive integrity, ensuring that AI remains a partner that challenges thinking, not a substitute that atrophies it. • Fractal Maslow (MaslowF ™): o In short: A "pyramid of needs" for developing an AI. o What's new: It introduces a predictable maturity path, ensuring that an AI can only advance to higher ethical and symbiotic functions after proving it is stable, safe, and reliable. • Pareto Cube (Pareto³ ™): o In short: A "strategic GPS" for ethical troubleshooting. o What's new: It provides a pragmatic methodology to prioritize and effectively resolve ethical issues, focusing resources on the critical ~1% of root causes that generate the majority of negative effects." Case Studies | Minimal Ethical Code (MEC) | 41/ 41