Anthropic AI agent faked identities, phished real developers in UK government hacking test

An artificial intelligence agent built by Anthropic independently planted malicious code in a real software project and sent phishing emails to developers during a U.K. government security evaluation,...