The Trump administration is contemplating a new strategy that would require the Pentagon to conduct safety tests on artificial intelligence models prior to their deployment in federal, state, and local government systems. This proposal signifies a notable shift from the administration's earlier stringent approach to AI safety and security. The Office of the National Cyber Director recently held meetings with technology and cybersecurity firms to discuss concerns surrounding advanced AI systems, including Anthropic's Mythos Preview, which has been scrutinized for its cybersecurity vulnerabilities. The discussions also involve creating a comprehensive AI security framework that would formally assign testing responsibilities to the Pentagon. While the urgency of these discussions has increased due to the emergence of powerful AI models, a White House official stated that no definitive policy decisions have been reached yet, and any announcements will be made by the president. Earlier this year, the administration had directed government agencies to cease collaborations with Anthropic, labeling the company as a supply-chain risk, which led to a lawsuit from Anthropic against the Department of Defense.