The Trump administration is preparing an executive order on artificial intelligence that would require developers to hand over new frontier models to the government weeks before public release, according to multiple sources briefed on the plans.
The order could arrive as soon as this week and represents a significant shift in the White House's approach to AI regulation. The administration had previously signaled a hands-off stance toward the technology, but recent advances in cyber-capable AI systems have prompted officials to reconsider.
Under the framework being drafted, AI companies would operate on a voluntary basis but would be expected to provide the government access to advanced models at least 90 days before launch. The arrangement would also extend access to certain critical infrastructure providers in banking, healthcare, and other sensitive sectors.
The order divides into two main components. The first focuses on traditional cybersecurity, aiming to shore up defenses at the Pentagon and other national security agencies while encouraging information sharing between the tech industry and government about breaches and threats. The second section targets what officials call "covered frontier models," establishing a government review process to assess which AI systems qualify for scrutiny and examination prior to their release.
Details remain murky about which agencies would oversee the reviews. Both national security and civilian government bodies appear to have roles in enforcing the arrangement, though the specific division of labor has not been finalized.
The proposal falls short of what some hardliners in Congress and among AI safety advocates have demanded. Those voices have pushed for more restrictive licensing and licensing regimes that would give Washington direct control over AI development. Growing concerns about AI's potential risks have emboldened those arguing for tougher guardrails, even as business leaders warn that over-regulation could slow American innovation.
The internal debate over the order reflects genuine tension within the Trump administration on how aggressively to regulate the sector. Some advisors favor minimal intervention, while others argue recent breakthroughs in AI capabilities demand closer government monitoring. That conflict has slowed the drafting process and left questions about which provisions will ultimately make it into the final version.
A White House official declined to confirm specifics of the order, stating that "any policy announcement will come directly from the President" and cautioning against reading too much into draft language.
Author James Rodriguez: "This is the Trump administration trying to have it both ways, signaling toughness on AI while avoiding the kind of hard restrictions that would annoy Silicon Valley donors."
Comments