OpenAI reports disrupting two AI-enabled "false front" influence operations

OpenAI said it had banned two covert influence operations — one from Russia, one from Iran — that used its models alongside more conventional tactics to build “false front” entities for laundering geopolitical messaging to target audiences. The Iranian operation ran a stable of seven fabricated “journalist” personas that pitched long-form articles to small and medium outlets worldwide and generated batches of social media comments, mostly about the US–Iran war. The Russian operation, which OpenAI called “Dark Clark,” co-opted unwitting people in Latin America to front what presented itself as an independent think tank, and separately fabricated “leaked” documents and audio scripts that circulated online.

Both operations used AI heavily to draft internal reports on their own effectiveness, employing what OpenAI described as questionable or outright deceitful methodology to exaggerate their results to their own backers. Using its IO Breakout Scale, which rates disrupted operations from 1 to 6, OpenAI assessed the Russian campaign as Category 5 — the first time it has rated a disrupted operation that high since it began publishing these reports — and the Iranian campaign as Category 4. Both reportedly placed material in mainstream media outlets rather than relying solely on social-media distribution, which OpenAI said was consistent with a pattern across the 30 covert influence operations it has exposed since 2024: operations that land content in real outlets tend to have the greatest reach.

OpenAI drew a direct comparison to pre-AI influence campaigns, noting the Iranian personas resembled “Alice Donovan,” a fabricated journalist tied to Russian military intelligence active in 2016–17, and the Russian operation’s front resembled “PeaceData,” a fake outlet run by individuals linked to the Internet Research Agency in 2020. The report is the latest in a recurring series OpenAI has published since 2024 on state-linked misuse of its models, following an earlier 2026 disclosure of a different Russian front group; OpenAI said its goal in publishing such reports is to make further disruption easier and continued operation of exposed fronts harder.