Anthropic Disrupts Three Iranian Propaganda Operations Using Claude AI

On Thursday, Anthropic, the U.S.‑based artificial‑intelligence company, announced that it had shut down three influence operations linked to Iranian state‑aligned institutions that had been using its Claude language model to design, produce and distribute propaganda.

The disrupted accounts were tied to the Islamic Culture and Communications Organization (ICCO) under the Ministry of Culture and Islamic Guidance, the Islamic Propaganda Office of Khorasan Razavi, and the Bina Cultural Observatory of the Islamic Propaganda Organization. According to Anthropic’s threat‑intelligence report, the actors framed their work as part of a “soft war” or “cognitive warfare” program aimed at shaping public opinion inside Iran and abroad.

The report explains that the groups treated Claude as a central administrative and operational tool. They built campaign plans, doctrine manuals, persona systems, target databases and ministerial planning documents that would normally require a fully staffed program office. The actors also used the model to generate content in Farsi, Arabic, Urdu, Malay, Spanish and English, with plans to expand into 20 languages.

While the accounts deliberately concealed their origins—using virtual private networks and foreign phone numbers to register and verify access to Claude—they repeatedly disclosed their locations, institutions and roles in conversations with the model.

In addition to content creation, the operations employed Claude to launder attribution. The model produced posts that appeared to come from foreign writers or independent media outlets and crafted hashtag campaigns designed to look organic. Distribution channels included Iranian platforms such as Eitaa, Bale and Rubika, as well as global social‑media sites X, Instagram, Telegram, TikTok and YouTube. Some material was also observed being shared through channels sympathetic to the Revolutionary Guard Corps. One operation produced ICCO‑branded ministerial documents outlining a nine‑part international influence portfolio and organizational plans for the funeral of Iran’s Supreme Leader, Ali Khamenei, following his death in the early days of the 2026 war.

Another operation, based in Mashhad, ran a content factory called “Manjanegh” or “Catapult.” The network used dozens of activists to repackage public reporting from Iranian security services under different personas without links to those services, and it allegedly amplified the material through paid campaigns across more than 100 Iranian channels, including IRGC‑linked outlets. An actor linked to the Bina Cultural Observatory generated messaging in the official voice of an IRGC spokesperson. During a campaign tied to the 2026 U.S.–Israel–Iran war, the network falsely attributed claims to Western research institutions such as CSIS, Brookings and RAND to give state‑backed messages additional credibility. The report also notes that the networks targeted international officials, Iranian opposition figures and the Baha’I community, using AI‑generated personas and coordinated distribution tactics to make state‑backed messaging appear independent.

Following the discovery, Anthropic identified and banned the accounts associated with all three operations, cutting off their access to Claude. The company also shared information about the activity with industry and research partners. Anthropic’s action is part of a broader pattern of the company’s efforts to mitigate misuse of its models, which have previously been used for missile guidance, biological weapons research and surveillance operations.

In the immediate term, Anthropic will continue to monitor for similar misuse and will maintain its partnership network to detect and disrupt future operations. The company’s report adds to a growing body of evidence that state actors are increasingly leveraging commercial AI systems to conduct sophisticated propaganda and information‑warfare campaigns.