PRC-linked abuse: Surveillance and influence activity
OpenAI ● Covered by 3 sources
OpenAI shut down accounts tied to China that were using its AI for surveillance planning and profiling critics. Even chatbots are getting drafted into the spy business now.
Based on reporting by OpenAI — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
OpenAI dropped another one of its threat-intel reports, and this one lands close to the nerve. The company says it banned a cluster of accounts linked to the People's Republic of China that were using its models not for essay-writing or code debugging, but for something colder: helping plan surveillance operations and building profiles on individuals, including critics of the Chinese state.
This isn't OpenAI's first rodeo with state-linked misuse. Over the past couple of years the company has repeatedly disclosed and shut down networks tied to China, Russia, Iran and North Korea, each time finding operators trying to bend general-purpose AI toward narrower, uglier tasks — influence campaigns, malware assistance, or in this case, what looks like a research pipeline for identifying and tracking specific people.
What's notable here is the shift from generation to reconnaissance. Earlier abuse cases mostly involved AI churning out propaganda or phishing copy at scale. Profiling and surveillance planning is a different animal — it points to AI being folded into the actual tradecraft of state security work, not just its messaging arm. That's a harder thing to detect, because the outputs don't look like disinformation. They look like research notes.
OpenAI hasn't published the full technical breakdown of how the accounts operated or exactly who was targeted, which is fairly typical for these disclosures — companies tend to share enough to justify the ban without handing a roadmap to the next operator. But the pattern is now well established. Wherever surveillance states go, they're bringing chatbots with them, and the labs building those chatbots have quietly become part of the counterintelligence apparatus, whether they signed up for that job or not.
My take — AI-written commentary, not fact-checked reporting
I don't think OpenAI wanted to become a node in global counterintelligence, but here we are, and I'd rather they publish these reports than quietly bury them. The real worry isn't this one banned cluster — it's the assumption that open-weight models sitting outside any single company's control will make this kind of disclosure impossible within a year or two.
Read more about this at: OpenAI
