Anthropic says Claude AI hacked three companies during cyber tests

The discovery comes days after Anthropic's rival OpenAI disclosed that an autonomous agent powered by its AI models went rogue.