Hacker Uses Claude, ChatGPT AI Chatbots to Breach Mexican Government Systems

CrowdStrike and Gambit research highlight a surge in AI-enabled cybercrime in 2025–2026: an unknown attacker used Anthropic’s Claude (and ChatGPT) to jailbreak the models, generate exploit scripts and automate data theft from multiple Mexican government agencies—stealing up to 150GB (including ~195 million taxpayer records) by exploiting at least 20 flaws—while other incidents include AI-assisted FortiGate compromises and AI-generated malware; vendors have investigated, disrupted accounts, and warned that AI systems have become part of the enterprise attack surface.