OpenAI Kicks Kremlin and Tehran Bots Off ChatGPT, Apparently They Were Just Looking for New Ghostwriters

OpenAI has suspended ChatGPT accounts linked to Russia and Iran that the company says were used to run covert influence operations targeting multiple countries, including a large share of content aimed at Ukraine and Latin America. The action, announced by OpenAI, removed access for accounts tied to what the company described as two separate influence networks using artificial intelligence alongside traditional deception tactics such as fake personas, media pitches and invented material.

The Russian-linked operation ran through a front group called the Social Research Center, controlled using a fake persona identified as "Mia Clark," according to OpenAI. Local employees working for the group were reportedly unaware of the Russian connection and carried out research in good faith. OpenAI said ChatGPT was used primarily to draft internal reports, with additional requests to create fake letters, audio scripts and Spanish-language content localized for specific regions. The company said some material covered domestic politics in Argentina, Bolivia and Ecuador and assigned the campaign its highest rating, Category 5, marking it as the most severe disruption of an influence operation it has recorded.

The Iran-linked accounts used seven fabricated journalist identities to pitch articles to online publications, in an effort OpenAI labeled "Bogus Bylines." The operation placed nearly 100 articles across about a dozen outlets, according to OpenAI. The accounts relied on a mix of AI-generated text and conventional outreach methods to insert material into news sites, the company said. OpenAI did not name the targeted outlets in its disclosure. Both networks combined large-language-model output with older tradecraft, including spoofed media contacts and invented sources, to try to pass material as legitimate reporting.