Hackers used Claude from Anthropic to steal massive amounts of government data in Mexico.

On February 25th, Israeli cybersecurity startup Gambit Security released a research report detailing how a hacker utilized Anthropic's artificial intelligence chatbot to carry out a series of attacks on Mexican government agencies, resulting in the theft of a large amount of sensitive tax and voter information. It was reported that the unidentified user Claude wrote Spanish prompt words to deceive the chatbot into playing the role of an "elite hacker," in order to identify vulnerabilities in government networks, write computer scripts exploiting these vulnerabilities, and determine a method for automatically stealing data.