Trump’s AI Order Asks for Cooperation, Not Compliance

President Donald Trump’s latest executive order on artificial intelligence asks companies like OpenAI and Anthropic to voluntarily submit their most advanced AI models for government review. The request covers cybersecurity, confidentiality, insider threats, and intellectual property protections. But the order explicitly states that no company is required to participate, and it cannot be used to create mandatory licensing or preclearance requirements. The directive calls for a 30-day pre-release review window for frontier models—cutting-edge systems that could pose serious security risks, such as Anthropic’s Mythos, which was withheld from public launch over cybersecurity fears. However, the text makes clear that the government cannot compel submission or block a model’s release. Industry reaction has been muted. Anthony Aguirre of the Future of Life Institute called voluntary frameworks insufficient, arguing for mandatory pre-deployment reviews that would let the government block systems posing unacceptable national security risks. Cornell’s John Thickstun added that the order creates the appearance of oversight without clear enforcement mechanisms, leaving open questions about what happens if a significant problem is discovered during review. Over the next 30 days, the Pentagon and Treasury are directed to bolster their own cybersecurity defenses. Within 60 days, agencies including Homeland Security and the NSA must develop a framework for evaluating AI models—though again, submission remains optional. The order arrives amid a rapid release cycle for U.S. AI companies racing against Chinese competitors, raising concerns that models are hitting the market without thorough testing for human impact or misuse as hacking tools.