The Trump administration lifted export controls on Anthropic's most advanced AI models Tuesday, ending a weeks-long government blackout that had cut off public access to the most powerful AI tools ever built — and restoring America's edge in a race where foreign competitors are closing in fast.
This matters because the original restrictions, imposed June 12, didn't just block foreign adversaries — they locked out everyone, including paying American customers. Anthropic had to kill access to Mythos and Fable entirely rather than sort out who was allowed in. The government effectively shoved the most advanced AI on earth into a vault and sat on the key.
Now the vault is open again. Commerce Secretary Howard Lutnick said Anthropic agreed to three conditions: proactively detect and address security risks, work with the government on protocols and standards for current and future models, and report any malicious activity. Anthropic will begin restoring access Wednesday.
Here's the tell: Anthropic had already pledged to do all of this voluntarily, months before the export rule existed. TechCrunch reported that cybersecurity experts saw the restrictions not as a security fix but as leverage — a way for the Trump administration to pressure Anthropic after its executives publicly criticized how the government might use the technology. The Guardian noted the controls were imposed on "national security grounds," but HotAir reported that Anthropic said the government "did not provide specific details about its national security concern."
The models are genuinely powerful. HotAir reported that NSA Director Gen. Joshua Rudd told Sen. Mark Warner that Mythos "broke into almost all of our classified systems, not in weeks, but in hours." An AI researcher testing Mythos found it could autonomously create powerful break-in tools against Linux, the open-source code underpinning most of modern computing. Amazon's CEO reportedly warned the administration that its team found ways to bypass Fable's security guardrails.
But the competitive reality cut through the caution. TechCrunch reported that Asian AI companies are releasing models approaching Mythos-level capabilities — among them Fugu and Tulonfeng. You can't throttle American innovation while the rest of the world catches up and expect to stay ahead.
The deal isn't clean liberty. WIRED reported that Anthropic leaders are still at odds with the White House over Fable's risk, and that the White House is "over" CEO Dario Amodei — cofounder Tom Brown has replaced him in high-stakes meetings. Lutnick had already cleared Mythos for release only to White House-approved customers. OpenAI's GPT-5.6 was also limited to vetted partners at the government's request. OpenAI CEO Sam Altman put it plainly: extensive safety testing "is not a bad idea. I just don't like the idea of the government picking the customers."
The administration is making up its AI rules in real time, as WIRED's headline put it. Trump killed Biden's AI executive order, then spent weeks improvising his own controls with no clear legal framework.
The question now: did Anthropic win its freedom, or just trade one set of chains for another? The models are back, but the government is still picking winners — and reserving the right to lock the vault again whenever it wants.