Overview
- The White House pulled the planned signing on May 21, 2026, after circulating a draft that would have created a voluntary federal review for the most advanced “frontier” AI models.
- The draft would have asked developers to submit models for agency review up to 90 days before release while explicitly disclaiming any mandatory licensing requirement.
- High-profile tech figures including David Sacks and, by some accounts, calls involving Elon Musk and Mark Zuckerberg pressed the administration in the hours before the ceremony, arguing the plan could slow company rollouts and hurt U.S. competitiveness.
- Security agencies pushed for a longer 90-day window and clearer agency roles while industry sought a much shorter 14-day timeline, leaving unresolved disputes over who would lead evaluations and how reviews would work in practice.
- With the order shelved and no new timetable set, the government is relying on ad hoc steps such as restricted access to Anthropic’s Mythos and voluntary CAISI arrangements, which leaves gaps in formal oversight and could shape whether safety or speed wins out next.