Trump’s AI order offers review time, keeps company leverage

voluntary 30-day – President Trump’s new executive order, signed Tuesday, asks AI companies to voluntarily submit advanced frontier models for government review 30 days before public release—but it does not make releases conditional on findings. The final framework leaves compan

For the third time in recent weeks. Washington is trying to reach for the steering wheel as the most powerful AI models approach public launch. President Trump’s new executive order. signed Tuesday. creates a government window to see what these frontier systems can do before they reach everyone else.

The problem is what happens after that window closes. The order does not condition public release on what agencies find. Instead, it asks AI companies to voluntarily submit advanced models for government review 30 days before they are released.

That “softer” approach marks a shift from an earlier White House direction. Just last month, a draft version would have required companies to provide a 90-day window. Tech industry executives opposed it, and the final order ended up with a tighter timeline.

The executive order also moved through internal debate before it landed. It was nearly signed. but after a phone call with former AI and crypto czar David Sacks. the order was put on hold. During another White House meeting on Monday. Sacks again stressed that longer wait times would stifle domestic development of AI models.

Free-market groups framed the outcome as a win for innovation. Wayne Crews, a Competitive Enterprise Institute fellow, said: “The administration deserves credit for recognizing that innovation, not precautionary regulation, is what made America the global leader in AI.”

The executive order carefully places the program in “voluntary” terms. It states that the government assessment program is voluntary for AI companies. and it also states that public release of new models is not conditional on the outcome of the assessments. Some tech policy analysts said the assessment effort therefore lands in a limited role—close enough for government agencies to review systems. but not empowered to slow them down.

Critics argued the voluntary structure flips the usual balance of power in favor of the companies being reviewed. Public Citizen. a consumer rights advocacy group. called it “a form of industry self-regulation.” The Future of Life Institute. a pro-regulation nonprofit. argued that highly capable models such as Anthropic’s Mythos require more than a “trust the companies” approach.

Jessica Ji. a senior research analyst at Georgetown’s Center for Security and Emerging Technology. warned that the framework does not match the kind of leadership the federal government has traditionally provided for public-private work and responsibilities usually handled by the government. including critical infrastructure.

“My impression is that it does not really establish the strong leadership that the federal government has traditionally had in terms of facilitating public-private partnerships and safeguarding responsibilities that have traditionally been left to the government like critical infrastructure. ” Ji told Fast Company.

The executive order does not spell out a detailed testing regime. Instead, it directs agencies to build the process and sets the framework for how the review is supposed to work.

It calls for the National Security Agency and other security-focused agencies to co-design the model assessment framework and determine cyber-risk thresholds. The focus is especially on advanced cyber capabilities and what qualifies as a frontier model for the review regime. The Treasury Department is tasked with establishing an AI cybersecurity clearinghouse to track the discovery and patching of software vulnerabilities exposed by new AI systems.

The 30-day window is meant to be used for “cyber capability evaluations. adversarial testing. and national-security review” of large AI models. the order says. The Commerce Department’s National Institute of Standards and Technology is set to play a key role. The executive order also points to the Center for AI Standards and Innovation—formerly the AI Safety Institute—which already evaluates frontier models.

Still, the question that hangs over the entire order is whether the government can truly build an independent assessment when companies control access, infrastructure, and the technical information needed to evaluate what the models can do.

Ji said the influence of AI companies is unlikely to stop at the executive order. She was watching how the dynamic could evolve when it comes to cybersecurity leadership.

“I’m personally very interested to see what this dynamic might look like in the future when it comes to who will lead on cybersecurity. ” Ji said. “Do the AI companies get to set the terms as they release models. especially with this kind of weakened 30-day voluntary commitment to give the government access ahead of time?”.

Her concerns were echoed by what is already happening outside the government framework. Many AI companies have begun creating their own versions of early access and pre-release testing.

Anthropic provided access to its Mythos model to a modest group of software and cybersecurity partners. On Tuesday, it extended access to 150 new partners across more than 15 countries. OpenAI, meanwhile, gave early access to its latest GPT-5.5 model to almost 200 trusted partners through its own early testing program. OpenAI also keeps a cybersecurity-focused version of the model available only to trusted partners.

Those company-led efforts may give outside experts a view of the most capable systems before wide release. But they also underline the tension built into the executive order itself: the government is asked to review systems in a 30-day window while the companies already hold much of the practical access and technical context.

It’s also unclear whether 30 days is enough time to properly evaluate advanced risks. Ji said the answer depends on whether evaluations can be performed with adequate capacity—and she doubts that the organizations best positioned to do those evaluations will necessarily have hands-on time for a full month.

“It depends on capacity to do evaluations. and I think the organizations best positioned to do those evaluations are the companies themselves. ” Ji said. “So obviously we have a bit of a transparency problem: There’s this huge information asymmetry between the companies and everybody else. including the government.”.

She added that the government may face practical hurdles beyond time, including finding the right AI research talent and compute resources, and managing access to the models while working out details of the partnership with AI companies.

“I think a month probably does not mean that testers will have 30 days hands-on with the model,” Ji said. “It might look more like two weeks after they work through all the paperwork. It’s hard to say whether 30 days is adequate.”

Trump executive order AI regulation frontier models cybersecurity National Security Agency Treasury cybersecurity clearinghouse NIST Center for AI Standards and Innovation Anthropic Mythos OpenAI GPT-5.5 David Sacks AI safety