Two influence operations used fake identities to get misleading material into legitimate media, according to an OpenAI investigation. The company says it banned the ChatGPT accounts involved. The operations were associated with Russia and Iran, and both used AI mainly to support internal reporting and adapt propaganda for different languages.
How the campaigns reached audiences
The operations focused on placing content in media outlets rather than building social media campaigns. That approach gave the material a route into established publications, while the fake identities obscured who was behind it.
The Iranian operation, called “Bogus Bylines,” used seven fake journalists to place nearly 100 articles about the US-Iran conflict in online outlets worldwide. The group also generated social media comments, but those received almost no traction.
The Russian operation, known as “Dark Clark,” sought to discredit Ukraine and destabilize local politics across Latin America. Its operators controlled a think tank through a fictitious persona and likely involved local staffers who did not know the true source of the effort.
Fabricated material prompted public responses
Dark Clark used fabricated audio files and documents. These materials led to fact-checks and official denials in Ecuador and Peru, bringing the operation beyond the circulation of online content and into public responses from politicians and others.
OpenAI rated the campaign category 5 out of 6 on its Breakout Scale. The rating reflected the fact that politicians reacted to the content. The company described it as the first category 5 case in two and a half years of reporting.
That assessment points to a distinction between producing misleading content and getting it to influence events. In this case, the fabricated material drew enough attention to prompt verification efforts and official statements. The source account does not say that the claims were accepted as true; it describes fact-checks and denials as responses.
AI supported the work behind the scenes
OpenAI’s account indicates that AI was not the main public-facing engine of either operation. Instead, the tools were used mainly for internal reporting and for adapting propaganda to different languages.
That supporting role matters because the campaigns relied on human-facing routes: false identities, media placements and, in the Russian case, a think tank persona. The Iranian operation’s comments on social media gained little attention, while articles in online outlets formed the bulk of its described publishing effort.
The reports also distinguish the two campaigns in their reach and effects. Bogus Bylines placed nearly 100 articles worldwide, but its social media comments had almost no traction. Dark Clark’s fabricated files produced fact-checks and official denials in two countries, and political reactions contributed to its high Breakout Scale rating.
What the investigation describes
Together, the cases show how influence operations can seek legitimacy by placing material in existing media environments. Fake bylines and fictitious organizational identities can make it harder for readers to see who is responsible, while multilingual adaptation can help content travel across audiences.
OpenAI had previously uncovered a different Russian campaign built around social media and a fake think tank. The newer cases described here instead emphasize planting content in legitimate outlets. The company’s findings offer a look at the methods it identified, including the use of fabricated material, false identities and AI-assisted internal work.