OpenAI Bans Russian, Iranian Fake News Operations

OpenAI banned a batch of ChatGPT accounts tied to a Russian operation called "Dark Clark" and an Iranian one called "Bogus Bylines," after its investigators found both groups using the chatbot to build fake personas and plant stories directly in legitimate news outlets. That's a departure from the usual playbook. Most influence campaigns OpenAI has caught so far leaned on social media bots and troll farms. This time the targets were real newsrooms in Latin America and online outlets worldwide, with fabricated audio, fake think tanks, and invented journalist bylines doing the work instead of viral posts.

The company's latest investigation report lays out how each operation used AI mostly behind the scenes, drafting internal talking points, translating propaganda for different markets, and occasionally generating social comments that barely moved the needle. The real damage came from content that made it past editors and into circulation, forcing fact-checkers and government officials in at least two countries to respond publicly. OpenAI tracks the fallout using its own Breakout Scale, a six-point system for measuring how far a disinformation campaign spreads before it's caught.