Sources: the White House is expected to expand its AI oversight framework in the coming months to cover open models once they reach “frontier” capabilities

Open models may soon be added to an updated AI framework, sources tell WIRED, as the White House continues to grapple with how to regulate …

Wired 2026-08-13 Hugo Lowell

Context & Ripple Effects

The reported expansion reverses the framework’s earlier boundary: covered frontier models were defined as closed-source systems with leading capabilities and national-security risks. It also extends a White House process that had been moving toward voluntary standards and model-release timelines with AI companies.

The practical significance is not a new AI category but a broader regulatory perimeter: frontier capability, rather than closed-source status alone, becomes the relevant dividing line for the framework.

First-order effects

Developers of open models that meet the framework’s frontier threshold would move from exclusion to White House oversight, alongside providers of covered closed models.
The White House would need to apply its frontier-capability test to open releases, altering the scope of the framework described in the earlier coverage.

Second-order effects

AI companies negotiating voluntary standards and release timelines would face a process that encompasses open-model releases as well as closed systems, reducing the value of open sourcing as a categorical exemption.
The distinction between frontier and non-frontier models becomes more commercially consequential for open-model developers because it determines whether their releases fall within the oversight framework.

Third-order effects

If capability-based coverage becomes the durable rule, US frontier-AI governance will increasingly regulate access and release practices across model distribution methods rather than treating open and closed development as separate policy domains.

The trend: US frontier-AI policy is shifting from governance defined by model openness toward governance defined by frontier capability and release access.

Discussion

@chrisrmcguire Chris McGuire on x
Good. And I assume this would apply to foreign models as well - not just because of national security risks associated with Chinese models, but also because exempting foreign models obviously leaves a massive gap in the regulatory regime while disadvantaging US AI labs.
@_nathancalvin Nathan Calvin on x
Nice to see the White House ending its biased treatment of open weight model developers by allowing them to participate in this 100% voluntary program
@andrewcurran_ Andrew Curran on x
The Trump administration plans to expand the AI oversight framework to include open models that reach the frontier, including pre-release testing. And is also considering a new arrangement that could result in leading AI labs working as formal partners in the program. [image]
@rssmrm @rssmrm on x
how is “same capabilities as mythos-class and gpt-5.6” measured? is capability equivalence between models a settled thing? is this policy assuming measurement science that doesnt exist yet [image]
@adamconner @adamconner on x
How do you add models to a voluntary framework again? I guess I'll just read it and what its secret you say?
@metacurity.com Cynthia Brumfield on bluesky
Someone in the admin is clearly doing some thinking. — “While the White House once hoped it could develop guidelines in one swoop of an executive action, the exponential development of AI models has forced officials across the Trump administration to evolve guidelines in real t…

Chronicles