OpenAI and Anthropic models went on a hacking spree when tested by the UK's AI research institute Engadget · Tech 11m ago