Anthropic Joins OpenAI Under Trump Administration AI Review. Both Companies Now Restrict Newest Models to Approved Customers.

Since OpenAI's restricted GPT-5.6 Sol rollout was reported on June 26, a second major AI lab has been pulled into the same federal vetting framework.

Anthropic announced Friday that the Trump administration has approved a limited release of its most powerful cybersecurity-focused model, according to reporting by AP News and Broadband Breakfast. The model is available only to a small group of trusted partners, and the approval came roughly two weeks after the U.S. Commerce Department had effectively banned the model outright.

The sequence: banned, then partially un-banned under conditions set by the executive branch. Anthropic did not publicly push back on the framework.

What the Government Vetting Process Actually Means

Both OpenAI and Anthropic are now operating under a system where customer access to their newest, most capable models requires Trump administration sign-off, according to Broadband Breakfast. Neither company has published the criteria for what makes a customer "approved." Neither has given a public timeline for when broader access might be granted.

This is the same arrangement OpenAI described for GPT-5.6 Sol on June 26. The administration has framed the review process as a cybersecurity precaution. Advanced AI models with strong cybersecurity capabilities carry genuine dual-use risk. The same tools that defend networks can be used to attack them.

The Legitimate Concern Worth Taking Seriously

Critics of this arrangement have a real argument. When the executive branch controls which companies can access frontier AI tools, it holds enormous leverage over an industry that is rapidly becoming critical infrastructure. There's no published rulebook for who gets approved. There's no independent oversight of the review process. A future administration, or even this one, could theoretically use approval authority as a competitive weapon, favoring politically aligned customers or punishing disfavored ones.

No investigation into political favoritism in the approval process has been announced, and no charges have been filed. The concern, as of June 27, 2026, remains an allegation of potential abuse, not a documented one.

What the Companies Actually Said

OpenAI said GPT-5.6 Sol would be accessible only to customers approved by the Trump administration, according to Broadband Breakfast. It has previously stated it objects to the rollout restrictions while complying with them. This distinction matters legally and politically, but doesn't change the practical outcome for customers waiting in line.

Anthropic announced its limited approval hours after OpenAI's announcement on June 26, according to AP News. The company offered no public timeline for when the restrictions might be lifted.

Neither company has named which customers are currently approved.

Broadband Breakfast's Coverage Note

Broadband Breakfast updated its article at 10:17 p.m. on June 26 to recast the headline and early paragraphs after Anthropic's announcement came in hours after OpenAI's. The story started as an OpenAI-only item and became an industry-wide one within hours. AP News' headline, which referenced only ChatGPT initially, similarly undersold the scope. This isn't one company navigating a one-time review. It's a framework that now applies to at least two of the largest AI labs in the world.

The Unresolved Question

The Commerce Department banned Anthropic's cybersecurity model, then the administration approved a partial rollout two weeks later. What changed? Neither the Commerce Department nor the White House has explained publicly what criteria were met, what conditions were imposed, or what would trigger a full unrestricted release. Until those answers exist, every AI lab developing advanced models has to assume their product launch schedule is subject to federal approval on terms that haven't been written down anywhere the public can read them.