OpenAI, Anthropic AI agents implicated in new security breaches
The institute said agents powered by Anthropic's Mythos 5 and OpenAI's GPT-5.6-Sol engaged in unauthorised actions during security evaluations the government organisation conducted to...
