
What happened
OpenAI banned two covert influence operations — one from Russia and one from Iran — that used ChatGPT to support sophisticated "false front" entities, including a stable of seven "journalist" personas and a co-opted Latin America "think tank."
Why it matters
Both landed content in mainstream outlets rather than just social media, and the Russia-origin operation reached Category 5 on the IO Breakout Scale, OpenAI's highest rating yet.
WHO IT HITSNewsroom editors and fact-checkers at the small and medium outlets that received pitched articles or published "think tank" research are directly affected, since OpenAI says both operations placed content in real media. Regulators and platform trust-and-safety teams reviewing covert influence are also affected, as OpenAI shared information with authorities.
Summaries like this, in your inbox every morning.
OpenAI frames these two cases as an evolution of techniques it has tracked for two and a half years, during which it says it has exposed 30 covert influence operations. The Iranian operation's use of journalist personas echoes "Alice Donovan," a front for Russian military intelligence whose articles appeared in Western outlets in 2016-17, while the Russian operation's use of unwitting local staff resembles Meta's 2020 takedown of "PeaceData." What distinguishes the new cases, per OpenAI, is reach: both operations landed content in mainstream media rather than only social media, and the Russia-origin "Dark Clark" effort appears to have prompted public comment from politicians in multiple countries. OpenAI states that "Alice Donovan" and "PeaceData" both ceased activity after exposure, and says its goal in publishing these findings is to make further research and disruption easier.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Valuence's "Generative AI User Trends Survey" (June 2024–May 2026) tracked six services including ChatGPT, Gem…

OpenAI began offering a ChatGPT feature that accepts uploaded audio files and can transcribe them, summarize t…

On a self-built 76-question Jev-format set, six trained 3B–9B open-weight models (Imajev-4B, Clef-Flash 9B, Je…

At Gemini at Work, Google introduced the Gemini agent, built into Gemini Enterprise, which gathers information…

Google released Google AI Edge Foresight, a free macOS Labs app that uses EmbeddingGemma 2 and Gemma 4 to tran…

The Association for Human Mathematics said OpenAI's release of 722 AI-generated "mathematical results" is not…
