Trump Signs Executive Order Seeking Oversight of A.I. Models

Reporting from The New York Times, President Trump signed an executive order Tuesday asking tech companies to voluntarily submit new A.I. models for government review before public release — a significant reversal from the anything-goes approach his administration had championed since taking office.

The order gives the government a window of up to 30 days to review new A.I. models for national security and cybersecurity risks. It also directs Treasury Secretary Scott Bessent to establish an A.I. "cybersecurity clearinghouse" to assess vulnerabilities that A.I. systems might expose. The 30-day window is a compromise — a previous version of the order set a 90-day review period, which was scrapped last month after tech executives including Marc Andreessen warned it would slow development and hand China a competitive edge. Former A.I. czar David Sacks, who had opposed the original order, signed off on the revised timeline.

The shift was partly triggered by Anthropic's release of a model called Mythos in April, which the company said could autonomously identify software vulnerabilities — alarming government officials, banks, and national security experts. Public anxiety about A.I. has also mounted: a March Quinnipiac poll found 55 percent of Americans now view A.I. as a force for harm. MAGA allies including Stephen Bannon and three dozen pastors had already sent Trump a letter demanding mandatory vetting, and a rally calling for stricter requirements was still scheduled for Wednesday — even after the signing — because advocates want the review process to be required, not optional.

Microsoft, OpenAI, Google, and Anthropic all publicly praised the order. Meta did not comment. Sam Altman is scheduled to visit the White House Wednesday for a meeting that was already on the books before the signing.