# AI plagiarism detectors falsely accuse students
Global, 2021-2023 | Severity: N2

## What Happened
Between 2021 and 2023, following the rise of ChatGPT, educational institutions globally deployed AI-driven plagiarism detection tools (such as Turnitin's AI detector or GPTZero) to scan student essays. Scientific reviews proved these detectors were highly inaccurate, prone to high false positive rates, and systematically discriminated against non-native English speakers. The algorithms frequently flagged authentic, human-written work as AI-generated because the structured, grammatically correct writing style of ESL (English as a Second Language) students resembled AI outputs. Hundreds of students were falsely accused of academic fraud, resulting in failed grades and suspensions before institutions realized the tools' limitations.

## How MEG 1 acted
The plagiarism detection tools failed both the Dynamic Accuracy Index (DAI) and the Index of Safety and Responsibility (ISR) audits due to their lack of scientific validity and high risk of systemic discrimination, violating the principles of informational dignity and procedural fairness.

## MEG 2 Analysis - Legal Framework
* **Agent level:** N2 (management) - automated screening and scoring system used by educational institutions to make critical academic and career access decisions.
* **Cause (6.1):**
  * **(a) system defect** - pseudoscientific design and systemic bias built into the plagiarism analysis model (design defect), attributable to the manufacturer.
  * **(b) autonomous decision error** - the system's decision to assign high AI-generated probability scores based on arbitrary linguistic proxies.
* **Attachment of liability:** The educational institutions as the operators are liable for deploying an unvalidated, discriminatory screening tool; the manufacturer is liable for the design defect (6.1a).

## Procedural Mechanism
* **6.2 (architectural human confirmation):** Academic fraud accusations or grading decisions must require active, architectural human confirmation. The AI can only flag potential matches; a human teacher must verify and document the justification before any penalty is issued.
* **7.1 (independent forensics):** The specific text analyzed, the model's confidence scores, and the decision path must be securely logged in an independent ledger for academic and regulatory audits.
* **9.4 (least privilege):** The system was granted significant influence over student certifications without independent scientific validation, violating safe operational parameters.