On Tuesday, June 2, 2026, U.S. President Donald Trump signed an executive order establishing a voluntary framework that permits AI developers like OpenAI, Google, and Anthropic to provide the government with access to their most powerful AI models for up to 30 days before public release. This measure was prompted by concerns over Anthropic's Mythos model, which could expose critical vulnerabilities. The 30-day period is a compromise between an initial proposal of 90 days and the tech industry's preference for 14 days. The order also directs the Treasury, NSA, and CISA to create an "AI cybersecurity clearinghouse" in collaboration with industry to identify and fix software vulnerabilities. This action follows President Trump's rescission of a previous Biden-era AI oversight order and contrasts with the European Union's binding AI Act.