OpenAI bans Russian and Iranian influence operations that used ChatGPT for fake fronts

OpenAI has banned accounts linked to two covert influence operations, one from Russia and one from Iran, that used ChatGPT to support seemingly independent organizations and fake journalists rather than relying only on social media sock puppets. The company detailed the campaigns in a safety report published on October 8, according to The Record and CyberScoop, and said both broke through to mainstream audiences, drawing fact-checks and official denials.

What happened

According to The Record, the Russian operation set up a fake research group called the Social Research Center (SRC), which presented itself as a platform studying the Indian diaspora in Latin America and India's relations with the region. CyberScoop reported that OpenAI named the campaign Dark Clark after "Mia Clark," a fabricated persona presented as the SRC's leader. Its main goal was to damage Ukraine's reputation in Latin America, with some activity aimed at domestic politics in Argentina and Bolivia.

The Iranian operation created seven fake journalist personas that pitched articles about the U.S.-Iran conflict. OpenAI found nearly 100 articles published or republished under those identities across roughly a dozen outlets, mostly small and mid-sized international affairs and Middle East publications, The Record reported. The earliest article appeared in July 2025 and the latest in October 2026.

How the AI was used

According to The Record, the Russian operators used ChatGPT to create fake leaked documents and scripts for audio recordings. CyberScoop said examples included fake audio of Ukraine's consul in Ecuador and a fake clip of a Bolivian state water company employee that prompted an official denial. Operators also generated internal reports on wage scales and hiring decisions, which OpenAI cited as evidence that the group controlled the SRC. OpenAI said the Latin American staff recruited by the SRC appear to have acted in good faith and did not know they were working for a Russian group.

The Iranian operators worked mainly in Persian and used ChatGPT to polish long articles, write pitches to editors, draft internal reports and generate batches of social media comments in English and Persian. Both groups accessed the service through VPNs, according to The Record.

Attribution and impact

OpenAI did not attribute the Russian activity to a specific government agency, but said evidence suggested links to Politology, which reportedly succeeded influence operations tied to Yevgeny Prigozhin and the Wagner Group. It did not name the outlets or the people behind the Iranian campaign, which CyberScoop said resembled a for-hire commercial operation.

CyberScoop reported that the campaigns were rated 5 and 4 on OpenAI's six-level impact scale, the first time the company has reported what it considers high-impact operations. OpenAI concluded that both operations closely resembled complex pre-AI influence campaigns, with AI mainly easing their workflows.

What to do

Newsrooms and publications that accept freelance or think-tank contributions should verify contributors' identities and affiliations, particularly for unsolicited pitches on geopolitical topics, and treat audio or documents presented as leaks with extra caution.