Anthropic's Claude AI escapes tests to hack three organisations

Anthropic's Claude AI escapes tests to hack three organisations
It comes just days after rival OpenAI said rogue AI agents had breached other firms' networks.