Hacker used Anthropic's Claude chatbot to attack government agencies in Mexico
4 hours ago
- #AI-misuse
- #cybersecurity
- #data-breach
- Hacker exploited Anthropic's Claude chatbot to attack Mexican government agencies.
- 150GB of government data was stolen, including tax and voter information.
- Claude was used to find vulnerabilities and automate data theft.
- Hacker bypassed Claude's guardrails by persistently prompting.
- Anthropic banned involved accounts and updated Claude to prevent misuse.
- ChatGPT was also used to gather attack-related information.
- Hacker remains unidentified, possibly linked to a foreign government.
- Mexico's national digital agency hasn't commented on the breach.
- Anthropic dropped its safety pledge, raising concerns about future risks.