The federal government now decides which artificial intelligence models Americans are permitted to use, and the first test case just locked the public out of the most powerful tools on the market while Washington picks who gets access and who doesn't.

Commerce Secretary Howard Lutnick sent a letter to Anthropic Friday approving a "limited re-release" of its Mythos 5 cybersecurity model to roughly 100 hand-picked organizations — government agencies and private companies — while the consumer-facing model, Fable 5, remains dark. No timeline for public access. No appeals process. Just a bureaucrat's signature deciding what technology belongs in your hands.

Two weeks ago, Lutnick invoked export control authorities to force Anthropic to shut down both Mythos 5 and Fable 5 entirely. The stated rationale: foreign nationals work at Anthropic and at partner organizations, so export control rules meant the models had to go offline for everyone — including every American who had been using Fable 5. On Friday, Lutnick wrote that "appropriate safeguards are in place to permit certain trusted partners" to access Mythos 5, according to Semafor, which first obtained the letter. He cited "significant progress" in daily talks between the government and the company.

Notice what's missing: any mention of when, or whether, Fable 5 returns to the public. The letter is silent on the consumer model. People close to the talks say they're "moving toward" releasing it, Semafor reported, but the timeline is unclear. European officials and U.S. allies have expressed frustration at their new dependence on Washington's gatekeeping.

The same day, OpenAI released its GPT-5.6 Sol model — but only to about 20 customers approved by the Trump administration. CEO Sam Altman called the staggered debut "bad news," acknowledging the company had planned a wider, open-access launch. OpenAI said it views the testing period as temporary, adding: "We don't believe this kind of government access process should become the long-term default."

The mainstream outlets frame this as responsible oversight. AP called it "unprecedented government vetting of AI products for cybersecurity risks." NBC highlighted that the administration was "not convinced that Anthropic's leadership understood the severity of their worries."

What they bury: the surveillance angle. The Daily Caller reported that Anthropic previously refused to allow the U.S. government to use its models for domestic surveillance — and got placed on a supply chain blacklist as a result. In January, the Department of War said it would only contract with AI companies that agreed to "any lawful use" and would remove safeguards on surveillance and autonomous weapons. Anthropic called for greater oversight in June but claimed the shutdown order went too far.

"We disagree that the finding of a narrow potential jailbreak should be the cause for recalling a commercial model deployed to hundreds of millions of people," Anthropic wrote. "If this standard was employed across the industry, we believe it would essentially halt all new model deployments for all frontier model providers."

They have a point. Trump's own adviser David Sacks acknowledged on a podcast that Anthropic's CEO Dario Amodei "spiked the cortisol level" in Washington by describing Mythos as a cyber weapon. So the company that went to Washington sounding alarms got punished for it — and now the regulatory framework is being built "on the fly," as Semafor put it, with no input from the public.

The real question isn't whether AI poses risks. It's who gets to decide whether you're trusted enough to use it. Right now, the answer is: not you.