President Trump signed an executive order Monday that creates a voluntary framework for reviewing advanced artificial intelligence models and expands AI-powered cybersecurity efforts. The order was delayed for months as administration officials worked through concerns tied to China's rapid AI development, according to sources familiar with the process.
Voluntary Review Framework for Advanced Models
The order establishes a process where AI developers can submit their most advanced models for government review, but submission is not mandatory. Companies that choose to participate will get feedback on safety, security, and potential national security risks. The framework is designed to cover so-called frontier models — the largest and most capable systems that pose the greatest unknowns.
Administration officials described the system as a way to stay ahead of risks without slowing down innovation. The review will be carried out by a newly designated body within the federal government, though the exact agency lead has not been publicly named.
Expanding AI in Cybersecurity
The order also directs federal agencies to ramp up their use of AI to defend government networks. That includes faster threat detection, automated responses to common attacks, and better coordination between civilian and defense cyber teams. The cybersecurity expansion builds on existing pilot programs that have used machine learning to spot unusual network activity.
Officials said the expanded effort will focus on protecting critical infrastructure like power grids, water systems, and financial networks. Private-sector partners in those industries will be encouraged to adopt similar AI tools, though the order stops short of requiring them.
China Factor Behind the Delay
The executive order had been in the works for roughly a year but stalled as the administration weighed how China's own AI push might affect the rules. Beijing has poured resources into AI development and has shown little interest in voluntary restraints. That made some White House advisors worry that a U.S. framework could hand an advantage to Chinese firms if American companies were effectively slowed down by paperwork while Chinese rivals raced ahead.
Negotiations inside the administration dragged on for months. Proponents argued that the voluntary system avoids heavy-handed regulation while still giving the government visibility into the most powerful models. Skeptics wanted mandatory reporting or outright restrictions on exports of AI technology to China.
The final version includes no new export controls, though the order does call for a review of existing policies on AI-related technology transfers. That review is due within 90 days.
What Comes Next
With the order signed, federal agencies now have 120 days to publish detailed guidance on how the voluntary review process will work. Companies building AI models that exceed a yet-to-be-defined threshold of computing power will be the first expected to participate.
The order also creates a new interagency council on AI that will meet quarterly to track risks and recommend adjustments. The council's first meeting is scheduled for next month, with a public report due by the end of the fiscal year.
