OpenAI Disrupts Two “False Front” Operations, Including Its First Category 5 Influence Operation

On October 8, 2026, OpenAI published its latest threat disclosure. It announced that it had banned two covert influence operations that used its models to support “false front” entities: one from Russia, which OpenAI calls “Dark Clark”, and one from Iran, called “Bogus Bylines”. A false front is an entity that hides who is really speaking. The operators create or borrow journalists, think tanks or news outlets that look independent, and use them to launder geopolitical and conflict-related messaging into a target audience. OpenAI notes that it has reported on threat actors for two and a half years, covering cyber attacks, covert influence operations, scams and other violations of its Usage Policies. The stated purpose is to inform regulators, industry peers and the wider public about how threat actors try to use AI, and about the vulnerabilities they try to exploit.

Start with the Iranian operation. “Bogus Bylines” kept a stable of seven “journalist” personas and used them to pitch long-form articles to small and medium online outlets around the world. An editor who received such a pitch saw a freelance writer, not a state-linked operation. The operation did more than write articles. It also generated batches of social media comments, generally on topics related to the US-Iran war. It used AI to draft internal reports as well, and it relied on questionable or outright deceitful methods to exaggerate its operators’ effectiveness. OpenAI assesses it at Category 4 on the IO Breakout Scale, which means its content did not stay inside the operation’s own circle but reached real audiences. Now the Russian operation. “Dark Clark” targeted Latin America. OpenAI says the operation appears to have co-opted unwitting people in the region to run a “think tank” on the ground. Besides controlling that think tank, the operators created fake “leaked” documents and audio scripts, and OpenAI identified some of them being spread online. Of all the uses it made of AI, drafting internal reports was the most common, and the same pattern of inflated claims about effectiveness appears here. OpenAI rates this operation Category 5, the first operation at that level that it has disrupted since it began its reporting. The most striking point is how closely both operations resembled complex influence operations of the pre-AI age. AI made some workflows easier, but the architecture was old. The Iranian personas bear a family resemblance to “Alice Donovan”, a fake journalist fronting for Russian military intelligence, whose articles ran in a range of Western outlets in 2016 and 2017. In 2020, individuals associated with past activity by the Russian Internet Research Agency ran a fake “news” outlet called “PeaceData”, which co-opted unwitting journalists around the world into writing for it. In other words, the false front is not an invention of the AI era. What AI changes is the cost of staffing it and the language skill it demands.

The report also offers a pattern drawn from data. Since early 2024, OpenAI has exposed 30 covert influence operations. It sorted them in a matrix by primary distribution method (social media, an operation-run website, or external publications) and by score on the Breakout Scale. Operations that use more than one method are classified by the method that appears to be the core of the operation. The finding is that operations which try to land content in real media outlets, rather than relying on fake social media distribution, tend to have the highest potential reach and impact. Both new cases did this, although not all the content they placed was generated by OpenAI’s models. That explains why they rank above most of the earlier operations, which stayed on social platforms.

From a defensive point of view, AI gives such operations more scale, efficiency, linguistic fluency and editorial ability. Operators can use these gains to exploit unsuspecting victims, such as employees or editors, and to put content in front of audiences who do not know who was behind it or what motivated it. Our reading is that newsrooms should treat identity checks on new contributors, and background checks on unfamiliar think tanks, as routine work rather than an exception. OpenAI also points out that the covert nature of false fronts makes them unusually vulnerable to responsible disclosure: both “Alice Donovan” and “PeaceData” stopped their activity once they were exposed. That is the aim of this report. It is meant to make further research and disruption easier, and to make continuing such operations harder.