NewsIncident
OpenAI bans Russian and Iranian operations that used ChatGPT to run fake research fronts and fake journalists
OpenAI reported on October 8 that it banned two covert influence operations running "false front" entities. "Dark Clark", from Russia, ran a Latin American "Social Research Center" through a fake persona while local staff, apparently unwitting, wrote most of its content; operators used ChatGPT to translate and proofread fakes and to write reports on their own work. Its fakes included emails to schools in Peru and Ecuador, fake audio about Ukraine's consul and about water cuts in La Paz, and a fake news video in Ecuador, which drew fact-checks and official denials; open-source researchers linked some of it to Prigozhin-linked networks. OpenAI rates it Category 5 on the Breakout Scale, the first it has disrupted at that level. "Bogus Bylines", from Iran and apparently a commercial for-hire campaign with an unidentified client, used seven fake Western journalists to place about 100 articles on the US-Iran conflict in about a dozen small and medium outlets from July 2025 to October 2026, with ChatGPT refining articles and pitch emails and writing batches of social media replies. OpenAI shared information with the relevant authorities.
Why it matters
The reach came from real outlets and real staff, not from the AI: a model makes fake bylines and pitches cheap and fluent, so editors should verify who a contributor is before publishing; one outlet that ran the articles says it is reviewing its submission process.