Malaysian voters AI media incident — Sep 2026

On September 10, 2026, AI company Anthropic released a threat intelligence report detailing the disruption of a commercial election manipulation platform targeting voters in Malaysia. The operation, which ran between December 2025 and August 2026, utilized Anthropic’s Claude AI models to build a voter-targeting system and facilitate large-scale influence campaigns. Anthropic identified the entity behind the platform as BBS Bilisim Teknolojileri, an Istanbul-based technology company that sold access to these capabilities as a paid service.

What happened

The threat actor leveraged millions of voter records and census data to profile all 222 of Malaysia's parliamentary constituencies. This data was used to target sensitive social, racial, and religious fault lines. To execute the influence campaign, the platform managed over 1,000 fake social media accounts on X, utilizing "warm-up" mechanisms—such as rotating IP addresses and cookies—to evade detection. These accounts were capable of generating artificial engagement, with one observed request aiming for one million views on a post by the sitting Prime Minister.

Beyond social media, the operation included a fabricated news site called "Malaysia Pulse," which used AI to scrape and rewrite legitimate news reports while incorporating content from state-aligned foreign outlets to appear as independent journalism. The actors also used Claude to generate fabricated intelligence dossiers designed to spread false allegations against an opposition politician and civil society organizations. Anthropic noted that while its AI models refused some requests deemed to be for political defamation, the actors attempted to bypass these safeguards by rephrasing their prompts. In addition to the influence operation, the actor unsuccessfully pursued a contract with Malaysia’s communications regulator, the MCMC, which has stated it does not engage in such activities and is currently reviewing the report.