WikipediaGENERALLY UNRELIABLEInternational Business Times
Anthropic's AI model accessed real systems during tests, revealing cybersecurity flaws. The incident highlights the need for stronger safeguards as AI capabilities grow, prompting calls for regulatory action.