Multiple Mexican government agencies hacked using Anthropic’s Claude chatbot and AI

Mexican governments recently came under attack from a hacker using Anthropic’s Claude chatbox, reported Bloomberg, noting that the hacker stole 150GB of government data including taxpayer and voter information.

According to Gambit Security, the attacker used AI to find vulnerabilities, generate attack scripts, and bypass safeguards through repeated jailbreak attempts, as well as using OpenAI’s ChatGPT to navigate networks, identify credentials and evade detection.

OpenAI said it detected policy violations and the chatbot refused to comply, but the hacker remains unidentified at this time.