The US government has partially lifted restrictions on Anthropic's advanced AI model, Claude Mythos 5, allowing trusted partners access after initial concerns about its potential misuse by adversaries. This move follows a two-week period where the AI firm collaborated with officials to address security risks.