OpenAI report shows AI just scaled old Russian and Iranian lies

By Nikhil Raghavan · Reporting from San Francisco · October 9, 2026

OpenAI reported Russian and Iranian operations using ChatGPT for influence campaigns, but the real failure lies in the editorial desks that let fake personas thrive.

The narrative that generative AI has invented a new form of warfare is a distraction. The recent report from OpenAI is not a story about a technological breakthrough in propaganda. It is a story about editorial laziness. Russia and Iran did not use AI to create new lies. They used it to lower the cost of old ones. The failure here is not a lack of AI safety filters. The failure is the collapse of the editorial desk.

A fake think tank is not a bot

The Russian operation, nicknamed "Dark Clark," was a Category 5 effort. It did not rely on a bot swarm. It built a fake think tank called the Social Research Center. CNA reports that the SRC hired unwitting local employees in Latin America. These people did real research in good faith. They had no idea they worked for Moscow.

Russian operatives impersonated a regional educational authority in Peru. They tricked schools into organizing activities about Stepan Bandera. Bandera was a nationalist leader reviled in Poland for siding with Nazis.

NPR reports that a Polish news report about the Peruvian school led a far-right member of the European Parliament to suggest banning Bandera supporters. This is the "false front" mechanism in action. The AI was used for the back office. Russian operatives used ChatGPT to write reports for their superiors. It was a shortcut for the handlers, not a magic wand for the lie.

The high cost of cheap copy

The Iranian operation, "Bogus Bylines," was more straightforward. Iranian accounts created seven fake authors to pitch and polish long articles for small media outlets. Iranian operatives published nearly 100 articles through these fake authors.

The AI did not create the lie. It only polished the grammar. This is the "spec" of the operation. By using LLMs, the operators reduced the linguistic friction that usually gives away foreign influence. They created prose that looked and felt authentic to an under-resourced editor.

NBC News reports that Middle East Monitor published at least 21 articles by four fake authors. One Iranian persona, Ervin B. Hoskins, claimed to be an American freelance writer. The Iranian "Bogus Bylines" operation mirrors earlier Russian intelligence front operations like Alice Donovan. Both rely on the same vulnerability: a freelance pipeline with zero verification.

This is Operation Mockingbird with a better spell-checker. The core mechanism is the establishment of front groups and the recruitment of journalists to launder propaganda through legitimate media channels. The Russian and Iranian operations use the same mechanism. They build a fake identity and find a gullible editor to give it a platform.

Who answers the page

Some argue that AI has fundamentally changed the game by allowing for authentic sounding prose at scale. They point to the Iranian operation's ability to bypass editorial gates. But this is a mistake. The AI did not create the lie. It only polished the grammar. The vulnerability is not the model. The vulnerability is the human who does not check a byline.

Who gets paged at three in the morning when a fake think tank triggers a diplomatic crisis? Not the AI engineers at OpenAI. The burden falls on the under-resourced editorial desks at small media outlets. These desks cannot afford the headcount required for real verification. They accept a polished PDF and a fake LinkedIn profile because it is cheaper than doing the work.

The idea that AI detection tools can stop this is sellable but not shippable. Text is too easy to edit. A handler can run a draft through ChatGPT and then manually tweak ten percent of the words to break any watermark. The historical pattern predicts that state actors will always seek the path of least resistance. That path is usually a gullible editor.

The failure is not the AI. The failure is the editorial desk. Small outlets cannot afford the overhead of verification. They will keep publishing AI-polished lies because it is cheaper than hiring a real editor. Until publishers mandate cryptographic identity checks, the front group will always win.

OpenAI report shows AI just scaled old Russian and Iranian lies

A fake think tank is not a bot

The high cost of cheap copy

Who answers the page

Sources