President Donald Trump has put off signing an executive order that would have established a process for the government to review AI models before they're released to the public. At the White House, Trump told reporters he wasn't satisfied with the order's language, saying "I didn't like certain aspects of it." He emphasized that the United States is currently ahead of China and other nations in AI development, and he doesn't want regulatory measures to interfere with that lead. The signing was reportedly delayed in part because not enough technology executives could arrange to travel to Washington, D.C. on short notice, robbing the administration of a high-profile photo opportunity.
The proposed executive order would have directed the Office of the National Cyber Director along with other federal agencies to create a framework for evaluating AI systems for security risks prior to their release. This push comes in direct response to concerns raised by recent AI developments, including Anthropic's Mythos model and OpenAI's GPT-5.5 Cyber, both of which demonstrated capabilities to rapidly identify and exploit security vulnerabilities. These releases raised alarms among security experts about the potential for advanced AI systems to be weaponized.
A major point of contention in the order's language involved a provision that would have required AI companies to share advanced models with government agencies between 14 and 90 days before their public launch. Trump indicated this timeline could have acted as an impediment to innovation, saying the language "could have been a blocker." The administration appears to be reconsidering the balance between ensuring AI safety and maintaining the competitive edge that U.S. companies currently hold over Chinese rivals in the global AI race.