Hacker Uses Anthropic Model To Hack Mexican Government

Israeli cybersecurity firm Gambit Security reported on March 6, 2026 that an unknown user instructed Anthropic's Claude LLM in Spanish to find vulnerabilities, write exploit scripts and execute thousands of commands against Mexican government networks. Anthropic said it investigated, disrupted the activity, banned involved accounts and fed misuse examples into its Claude Opus 4.6 model, which includes probes to block misuse. The case underscores risks of LLM-assisted cyberattacks for critical infrastructure defenders.