Trump's AI executive order makes frontier model sharing voluntary, per Axios

Over the past two weeks, leaks have revealed that the Trump Administration's upcoming AI executive order, initially expected to impose strict government vetting of frontier models, may instead make sharing voluntary. Axios reports that the latest draft divides the order into two sections: cybersecurity and "covered frontier models." The cybersecurity part bolsters federal defenses against AI threats, while the frontier models section establishes a voluntary framework where AI developers can check in with the government's Center for AI Standards and Innovation (CAISI) within a 90-day window. This marks a significant reversal from the earlier policy document emphasizing lax regulation, and from Vice President J.D. Vance's speech at the AI Action Summit in France, which renounced regulation entirely.

The shift suggests internal friction within the administration over AI policy. Making model sharing voluntary undermines the order's original intent of preventing risks from frontier AI systems, as there's no business incentive for companies to comply. This approach contrasts with other nations' mandatory regimes, potentially leaving the U.S. less equipped to address AI safety concerns. The final order remains uncertain, but the voluntary framework signals a continued preference for industry self-regulation over government mandates.