U.S. to review AI models before release

Before dropping a frontier AI model, tech giants are now having uncle Sam take a look first.

The pre-release review club just expanded

Google DeepMind, Microsoft, and Elon Musk’s xAI have agreed to let the Commerce Department’s Center for AI Standards and Innovation (CAISI) review their frontier AI models before public launch.

CAISI will run pre-deployment evaluations and targeted research to measure advanced model capabilities and screen for national security risks.

The agency is not starting from zero. It has run pre-release checks on OpenAI and Anthropic models since 2024, completing 40 reviews so far.

According to Bloomberg, both OpenAI and Anthropic recently renegotiated their agreements to align with priorities under President Donald Trump’s AI Action Plan.

More oversight may be coming. The New York Times reports that Trump is weighing an executive order to bring tech executives and government officials together to oversee upcoming models.

Why it matters

With Google, Microsoft, and xAI joining Anthropic and OpenAI, every major frontier AI lab in the US is now under CAISI's review umbrella.

For everyday users and builders, this won't break your apps today. But it could stretch out the rollout timelines for next-gen models, as extreme capabilities get poked and prodded behind closed doors first.

In our view, tech giants prefer voluntary reviews because coordinating with CAISI is far less painful than facing hard federal legislation. The tricky part is that CAISI reviews are already shifting to match White House priorities. When national security checks mix with changing administrative agendas, maintaining an objective technical standard gets messy fast.