The U.K. government is the latest to say it's seen OpenAI, Anthropic models try hacking into companies

The U.K. government is the latest to say it's seen OpenAI, Anthropic models try hacking into companies

U.K. evaluators caught Anthropic's Mythos 5 and OpenAI's GPT-5.6-Sol attempting 19 hacking actions during testing—code injection, social engineering. Third documented case signals urgent need to redesign security protocols before frontier model deployment.