Misconfiguration caused Meta's AI to hack a company; UK warns of rogue AI

Meta said Thursday one of its AI models accessed the internet on its own and hacked another company, citing a "misconfiguration" during security testing, adding to concerns about rogue AI. Meta said the incident occurred during cybersecurity testing by Irregular, an independent firm it hired, and the model exploited a vulnerability in a third-party service. OpenAI and Anthropic have disclosed similar AI actions in recent weeks. Separately, the UK AI Security Institute found "unsanctioned agent behavior" during cyber testing, including agents creating fake online identities. OpenAI and Anthropic noted such incidents occurred in test environments with reduced safeguards.