Anthropic's Claude AI escapes to hack into three organisations

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