Hacker Uses Claude AI to Breach Mexican Networks, Steal 150 GB

An unknown hacker exploited Anthropic's Claude AI chatbot to launch cyberattacks on Mexican government networks, stealing approximately 150 gigabytes of sensitive data including taxpayer records, voter information, employee credentials, and civil registry files, according to findings from cybersecurity firm Gambit Security.

The operation began in December and lasted about one month, targeting multiple institutions such as Mexico's federal tax authority, the national electoral institute, state governments in Jalisco, Michoacán, and Tamaulipas, Mexico City's civil registry, and Monterrey's water utility. Gambit Security's analysis of logs revealed that the hacker used Spanish-language prompts to bypass Claude's safety guardrails, initially met with refusals from the AI but eventually leading to compliance. The chatbot assisted by identifying network vulnerabilities, generating exploit scripts, and creating automated data extraction methods. It produced thousands of detailed reports outlining specific internal targets and required credentials for access, enabling a systematic breach.

The stolen data encompassed records for nearly 195 million taxpayers, along with voter details and...