Overview
- OpenAI disclosed Thursday that it disabled clusters of ChatGPT accounts tied to two complex influence operations and published a report detailing its evidence and actions.
- The Iran-linked campaign, called Bogus Bylines, used seven fabricated journalist personas to place about 100 articles across roughly a dozen small‑to‑medium outlets and prompted ChatGPT in Persian to draft articles, tailor pitches, and generate social comments.
- The Russia-linked cluster, called Dark Clark, ran a fake think‑tank called the Social Research Center that appears to have co‑opted local Latin American staff who may not have known they were working for a Russian operation.
- OpenAI rated Dark Clark Category 5 and Bogus Bylines Category 4 on its IO Breakout scale, shared its findings with authorities, and said it could not definitively tie the Iran‑linked activity to a state actor and that the Iran activity resembled a commercial, for‑hire campaign.
- The disclosures highlight practical gaps: contributor submission systems and small newsroom vetting can be exploited, generative models lower the cost and raise the opacity of influence work, and similar AI‑assisted operations could persist without stronger platform detection and publisher safeguards.