OpenAI bans Russian, Iranian AI influence ops using fake media fronts

One Russian operation reached the highest severity tier OpenAI has recorded since it began reporting disruptions.

OpenAI has banned two covert influence operations, one Russian and one Iranian, for using ChatGPT to support elaborate false-front entities. These were built to launder geopolitical messaging into target audiences. The Iranian operation ran seven fake journalist personas. They pitched long-form articles to small and medium outlets worldwide. The Russian operation, nicknamed 'Dark Clark' by OpenAI, appears to have co-opted unwitting people in Latin America. They staffed a fake think tank called the Social Research Center.

OpenAI assesses the Russian operation as Category 5 on the IO Breakout Scale, a 1-to-6 measure of influence-operation reach developed by Brookings. It is the first Category 5 operation the company has disrupted since it began this reporting two and a half years ago. The Iranian operation reached Category 4. Both landed content in mainstream media outlets rather than relying on social media distribution. OpenAI says this fits a pattern across the 30 covert influence operations it has exposed since early 2024: operations that place content in real outlets achieve greater reach than those confined to social posts.

The Russian operators mostly prompted in Russian and used VPNs to get around OpenAI's ban on access from Russia. They used ChatGPT chiefly to draft internal reports to an unnamed superior. These covered efforts to denigrate Ukraine, interfere in the domestic politics of Argentina and Bolivia, and manage the Social Research Center. OpenAI says the reports described specific fabricated stories. One was a false claim that Argentine President Javier Milei bought jeweled dog collars. Another was a paid anti-Milei graffiti campaign in Buenos Aires. Open-source researchers had previously attributed both to a Russian entity known as 'Politology' or 'La Compania,' a reported successor to the Wagner Group.

OpenAI says the operation appears unusual in one respect. The Social Research Center produced most of its content through genuinely co-opted local staff, who were unaware they worked for a Russian group. Some of the fakes it spread provoked fact checks and official denials in Ecuador. Others fed into public tensions between Ukraine and Poland.

The Iranian operation worked differently. It used ChatGPT to generate batches of social media comments on the US-Iran war, alongside its journalist personas' long-form pitches. OpenAI notes both operations resembled pre-AI influence campaigns. Examples include the Russian military intelligence front 'Alice Donovan,' active in 2016-17, and the 2020 fake outlet 'PeaceData'. Both of those ceased activity once exposed.

OpenAI says it is publishing details of the false fronts precisely because their covert nature makes them vulnerable to disclosure. Its aim, the company says, is to make further disruption easier and continued operation harder.

Written from the primary source above by The Neural Net’s automated desk (claude-sonnet-5, theneuralnet/pipeline@1) without human review. How this works. Corrections: check the source first — if we got it wrong, the source link is the record.