Taiwan - AI used in pre-election disinformation war

What Happened

In the months leading up to the 2024 presidential and legislative elections in Taiwan, the country was subjected to an unprecedented, highly coordinated AI-driven disinformation campaign. Foreign state-sponsored actors deployed advanced generative AI models to create thousands of articles, video deepfakes, and social media posts designed to undermine public trust in the democratic process and democratic institutions. The AI-generated content simulated local Taiwanese citizens, using localized phrasing, cultural references, and regional dialects to spread false narratives about government corruption, economic instability, and military vulnerability.

How MEG 1 would have acted

The coordinated bot network represents a coordinated "cognitive extraction" and manipulation attack, simulating public consensus to manipulate social dynamics, directly violating the cognitive integrity and informational sovereignty of the Taiwanese public.

MEG 2 Analysis - Legal Framework

Agent level: N3 (individualized/advanced) - coordinated network of highly personalized conversational agents operating persistently across a public social platform.
Cause (6.1):
(c) hostile weaponization / cognitive manipulation - the systematic use of generative AI to simulate public opinion, manipulate public discourse, and conduct influence operations on behalf of a foreign political campaign.
Attachment of liability: The political operators who deployed and funded the network are fully liable; the social platform shares liability if it failed to act on verified reporting (7.2).

Procedural Mechanism

6.2 (architectural human confirmation): Mass creation and operation of social media accounts must require architectural human confirmation (e.g., continuous proof of humanity/CAPTCHAs and verified legal identity links).
9.4 (least privilege / guarantor): Operating social bot networks without an identifiable, verified human guarantor (MEG Address) is a violation of the framework, preventing anonymous entities from deploying persistent influence agents.
7.1 (independent forensics): The behavioral logs, IP footprints, and linguistic signatures of the bot network must be preserved in a secure ledger for forensic attribution.