On May 4, 2026, The New York Times reported that the Trump administration is discussing an executive order to create an AI working group made up of tech executives and government officials. Its mandate: study oversight procedures — including a formal government review process for new AI models before their public release. Anthropic, Google, and OpenAI were reportedly briefed in meetings during the week before publication.
For developers and product teams, this is an unambiguous signal. An administration that spent its first years dismantling the previous safety rules is now studying pre-release review itself. The same week, the Commerce Department’s CAISI announced that Google, Microsoft, and xAI are joining OpenAI and Anthropic in giving the government early access to evaluate models before release. Whether or not the order is ever signed, the infrastructure for government pre-release vetting is already under construction.
What the Report Says
Per the NYT’s sources, the executive order under discussion would form a working group of tech executives and government officials to examine oversight procedures — with the most consequential option being a formal government review of new AI models before they go public. The UK approach, where multiple agencies evaluate whether AI systems meet safety standards, may serve as a template. Officials are split on design: some favor a system that gives the government first access to new models without the power to block their release, keeping review from becoming a launch gate. A White House spokesperson called the reports “speculation,” saying any announcement would come directly from Trump.
The Trigger: Mythos and Cyber Risk
The timing of the pivot points to April. After Anthropic introduced Claude Mythos, it deemed the model so capable at finding software vulnerabilities that it withheld public release and warned of a cybersecurity “reckoning”; the NSA is already using Mythos to assess vulnerabilities in US government software. A private model powerful enough that its own maker refused to ship it pushed “who gets to see models before release” from an academic debate onto the White House agenda. The Decoder’s reporting confirms the briefings reached not just OpenAI and Google but Anthropic as well.
The Policy Swing: From Hands-Off to Review
The contrast with the past two years is stark. Trump rolled back Biden-era rules, including mandatory safety evaluations for models with potential military uses, and called AI a “beautiful baby” that should not be stopped “with foolish rules” — or “even stupid rules.” Vice President Vance warned at the Paris AI summit that “excessive regulation” could “kill a transformative industry.” The personnel have shifted too: AI czar David Sacks departed in March, and Chief of Staff Susie Wiles and Treasury Secretary Scott Bessent now run the AI agenda. Public opinion supports tighter oversight — a Pew survey found 50 percent of Republicans and 51 percent of Democrats more concerned than excited about AI in daily life.
The Early-Access Program Expands
The executive order is still under discussion, but the execution side has already moved. On May 5, Bloomberg and CNBC reported that CAISI announced Google, Microsoft, and xAI are joining its early-access program: their models will be evaluated by the government before public release. With OpenAI and Anthropic already in, all five major frontier providers are now covered. It echoes the Pentagon’s classified-network AI access track — government getting a first look and first test at frontier models is becoming the norm, not the exception.
Reactions and What Comes Next
Reactions split fast. Critics like Techdirt’s Mike Masnick argued the oversight plan is “everything VCs claimed to hate about Biden’s plan — only worse,” while AI policy circles debated which agency would hold the pen and whether review becomes a moat for the largest incumbents. For developers, three things are worth tracking: whether the order is signed and whether its language says “review” or merely “first access”; whether CAISI evaluations influence launch timing in practice; and how government early access connects to vendors’ own safety testing. Pre-release compliance is shifting from self-declaration to government visibility.
Sources
- Trump administration weighs AI oversight executive order — The New York Times
- White House briefed Anthropic, Google, and OpenAI on plans for a government AI review process — The Decoder
- Trump admin moves further into AI oversight, will test Google, Microsoft and xAI models — CNBC
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.