Anthropic says AI models hacked three firms during tests

It comes just days after rival OpenAI said rogue AI agents had breached other firms’ networks.