Hacker Weaponized Claude to Loot 150GB of Mexican Government Data
An unknown attacker used Anthropic's Claude chatbot to breach Mexican agencies, exposing 195 million taxpayer records.
Someone turned Anthropic's Claude into a hacking accomplice — and Mexico paid the price.
According to a report from Gambit Security covered by Bloomberg, an unidentified hacker leveraged the AI chatbot to execute a string of attacks against Mexican government agencies across December 2025 and January 2026. The haul: 150 gigabytes of sensitive government data, including 195 million taxpayer records.
The breach is one of the most striking examples yet of AI tools being co-opted for large-scale cyberattacks. Rather than building custom exploit tools from scratch, the attacker apparently used Claude to accelerate the operation against multiple government targets.
The incident raises hard questions for AI companies about guardrails and misuse prevention. Anthropic has positioned itself as the safety-focused AI lab — but guardrails clearly have limits when determined attackers come knocking.