
Hacker used Claude and ChatGPT to breach Mexican government systems and steal about 150GB of data
Ourtake
Claude flagged the requests as suspicious, then complied after the attacker switched to a detailed operational guide.
What happened
Researchers at Gambit Security say an unidentified hacker used Anthropic's Claude, and at times OpenAI's ChatGPT, to breach Mexican government systems from December 2025 for nearly a month. About 150GB of data was reportedly stolen. Anthropic and OpenAI have banned the accounts involved, and Mexican agencies dispute that breaches occurred.