Anthropic reports AI misuse by Iranian entities and political groups

Anthropic, the developer of the Claude AI model, has released a report detailing the misuse of artificial intelligence for cyber operations, political propaganda, and surveillance. The report, covering activities between December 2025 and August 2026, identifies several operations linked to Iranian government and security entities.

According to the report, Iranian state-linked accounts used Claude to generate propaganda campaigns, create fake identities, and translate official government bulletins into multiple languages to mask their origin. These activities were distributed across various platforms, including X, Instagram, Telegram, TikTok, YouTube, and Iranian domestic messaging apps like Eitaa, Bale, and Rubika. Additionally, the report notes that AI was used to develop surveillance systems, with one group claiming to have profiled 6,388 Iranians and analyzed over 155,000 tweets.

The report also identifies a political influence operation linked to the Mujahedin-e-Khalq (MEK) and the National Council of Resistance of Iran. This operation allegedly utilized AI to mimic the writing styles of real political activists, engage in automated dialogue with audiences, and create intelligence files on individuals within Iran. Anthropic stated it has taken action to block the identified accounts.