OpenAI published its latest threat report on February 25, 2026, detailing how criminals and state-linked actors tried to weaponize ChatGPT for scams, propaganda, and surveillance. The report marks two years since OpenAI became the first AI lab to publish regular threat disclosures.
What Bad Actors Did
Three cases stand out. A user connected to Chinese law enforcement used ChatGPT to run smear campaigns targeting Japan’s prime minister and critics of the Chinese Communist Party, including generating fake sex scandal accusations against dissidents. A Cambodia-based scam network combined manual ChatGPT prompting with an automated AI chatbot to run romance fraud, walking targets from social media engagement toward fake investment schemes. A Russia-linked content farm used the model to translate and generate posts for distribution across social media platforms.
The report’s consistent finding across all cases: AI was one tool in a larger kit. Bad actors layered it on top of fake social media profiles, forged documents, and paid advertising. In none of the cases did AI-generated content prove to be the decisive factor in a campaign’s success.
Two Years of Takedowns
OpenAI started publishing these disruption reports in February 2024. By October 2025, the company had banned and reported over 40 networks for violating its usage policies. The February 2026 report continues that series, now covering cyberattack tooling, covert influence operations, and financial fraud. OpenAI calls for broader information sharing between AI companies, governments, and the security community to stay ahead of actors who, once banned, tend to move to less-monitored tools.