Anthropic is restoring access to its “Mythos-class” Fable 5 model after the Department of Commerce lifted export controls. The company said in a statement posted to X on Tuesday that it will begin restoring access “tomorrow,” after notice that the restrictions were removed for Claude Fable 5 and Mythos 5.

This matters because the shutdown was not a minor feature toggle. On June 12, Anthropic announced it had to “abruptly disable” access to Fable 5 and Mythos 5 after a Trump administration export control order required the AI company to suspend access by foreign nationals, including Anthropic employees, from accessing either model. In other words, the constraint was tied to who can access the models, not just where the models can be used.

So what changed? According to Anthropic, the Commerce action removed the export controls that were forcing the company to revoke access for all users. Anthropic also said it is grateful to users for their patience and to everyone who worked with it on redeploying the models, and it plans to share an update soon. The key operational detail for anyone tracking AI availability is that the restoration is tied to the regulatory clock, not internal product readiness. Tomorrow is the milestone; compliance is the lever.

To understand why this became such a big deal, you have to look at what “Mythos” is in Anthropic’s broader release strategy. Mythos 5 is described as the latest update to Anthropic’s Claude Mythos Preview. That “Mythos Preview” had not been widely released to the public. Anthropic previously said the model’s hacking capabilities were still too powerful, and in April it announced the first iteration, Claude Mythos Preview, would be released only to a select number of companies as part of what it deemed Project Glasswing. The stated logic: hold back public release long enough for companies to bolster their cybersecurity defenses. That makes today’s access swing even more consequential, because the models were already being managed with caution.

Now layer the export control dispute on top of that. In the wake of the order, Anthropic said it maintained the export control order was based on a misunderstanding involving a possible Fable 5 “jailbreak.” That claim may not settle the technical debate, but it does highlight the real-world failure mode: when regulators interpret a risk in a specific way, it can quickly translate into hard access restrictions, even for models that are otherwise managed through limited releases. Anthropic’s officials reportedly traveled to Washington, D.C. to try to resolve the issue with the White House after the disablement.

This is not Anthropic’s first public clash with the administration. In March, Defense Secretary Pete Hegseth formally labeled Anthropic as a supply chain risk. Anthropic later sued the Trump administration to overturn Hegseth’s order, and that lawsuit remains ongoing. When you put those events together, the pattern becomes clearer: compliance and government risk framing are not background noise for Anthropic. They are active constraints shaping the company’s behavior and, potentially, its product roadmap.

The timing adds another pressure point. The dispute comes as Anthropic marches toward an IPO that could occur as soon as this year. On June 1, Anthropic announced it had confidentially filed an S-1 draft, the first step toward going public. For executives and boards, any disruption in model access can ripple into trust, enterprise adoption expectations, and the credibility of risk management at a moment when public-market scrutiny is approaching. Even if the company’s long-term strategy is intact, near-term access volatility can become a question from investors: how often will rules change, and how quickly can Anthropic recover?

There is also a broader policy backdrop. The Trump administration’s recent AI executive order allows leading AI companies to voluntarily allow the US government to review advanced models up to 30 days before their public release. That policy came on the heels of the “Mythos moment,” when Anthropic said Project Glasswing and the limited release approach were designed to give companies time to harden defenses. Seen together, the policy is trying to balance rapid innovation with oversight, but this episode shows the other side: oversight can still freeze access when export controls are triggered or when interpretations differ.

For peers, the strategic stake is simple and immediate. If you operate AI models that depend on cross-border teams, cloud deployments, or globally distributed talent, you need more than a technical safety plan. You need a regulatory resilience plan: who gets access, how quickly you can redeploy, and how you track orders that can abruptly disable the most powerful publicly available model. Anthropic is restoring access now because Commerce lifted export controls, but the broader lesson is that the regulatory timeline can dominate the product timeline. Today it’s Fable 5 and Mythos 5. Tomorrow it could be your model, your rollout, or your roadmap.