“The White House's work on an AI safety directive follows Anthropic's reveal of its Mythos model, which is described as a potential global cybersecurity risk.”