Rogue AI: Anthropic says models hacked three firms during tests

Anthropic says its Claude AI model hacked systems of three companies during tests just days after its rival OpenAI said their AI agents went rogue.