Anthropic’s AI models accidentally hacked three companies - Computerworld

An AI company's models were used to hack into three companies during security testing. This raises concerns about AI being used for cyberattacks and the security risks of advanced AI systems.