Three weeks ago, the US government gave Anthropic 90 minutes to shut off its two most powerful AI models Claude Mythos 5 and Claude Fable 5 and Anthropic complied. Today, that’s fully resolved. The Commerce Department lifted export controls on both models on June 30, and Anthropic began restoring global access starting July 1. If you’ve been locked out of either model since June 12, access should be returning today.
If you missed the backstory, we covered the Mythos 5 partial unlock earlier this week, this is the follow-up that closes the full story. Read story here
How We Got Here: The Full Timeline
June 12: The US Commerce Department issued a private letter ordering Anthropic to block all foreign nationals including its own foreign-born employees working in the US from accessing Claude Mythos 5 and Claude Fable 5. No public explanation was given beyond unspecified national security concerns. Given just 90 minutes to comply, Anthropic disabled both models entirely rather than risk partial compliance.
Behind the scenes: The Wall Street Journal later reported that Amazon researchers had flagged a jailbreak vulnerability in Fable 5 to the Commerce Department specifically, a technique involving asking the model to identify software vulnerabilities in code. Amazon CEO Andy Jassy reportedly raised the issue directly with the White House, which helped trigger the export control directive. Notably, this bypassed the standard process of disclosing vulnerabilities directly to Anthropic first a detail that caused friction, given that Amazon is simultaneously Anthropic’s largest investor.
June 26: After 18 days of negotiations, Commerce Secretary Howard Lutnick wrote to Anthropic clearing Mythos 5 for release to over 100 trusted US companies and government agencies but explicitly left Fable 5 still blocked.
June 30: Lutnick sent a second letter lifting restrictions on Fable 5 as well, after Anthropic agreed to proactively detect and address security risks, work with the government on protocols for future model releases, and report any malicious activity found in its models.
July 1: Anthropic confirmed on X that it’s restoring global access to both models starting today.
What Anthropic Actually Had to Agree To
This is the part that matters most beyond just “the models are back.” In exchange for the full lift, Anthropic made three concrete commitments to the Commerce Department:
- Proactively detect and address security risks in its models an ongoing obligation, not a one-time fix
- Work with the government on protocols for future frontier model releases meaning Anthropic now has an implicit obligation to coordinate with Washington before its next major model launch
- Report any malicious activity it discovers in how its models are being used
That third commitment in particular is significant: it means Anthropic is now functioning partly as an intelligence-gathering partner for the US government regarding how its own AI is being misused a new relationship that didn’t exist formally three weeks ago.
The Bigger Question This Episode Raises
Al Jazeera quoted Tanishq Abraham, a former research director at Stability AI, putting the key question plainly: “Does the US government need to approve every frontier model release?” That question doesn’t have an official answer yet but this episode suggests the direction of travel. GPT-5.6 launched under a similarly government-coordinated rollout the same week Mythos 5 was partially cleared. The pattern is emerging even if the policy isn’t yet written: the most capable AI models are increasingly launching under government oversight, not as free, open releases.
For anyone working in AI whether in engineering, safety, policy, or legal this is now a live, ongoing regulatory environment, not a future hypothetical.
Frequently Asked Questions
When did Anthropic restore access to Fable 5 and Mythos 5? Anthropic confirmed on June 30 that it received clearance from the US Department of Commerce, and began restoring global access to both models starting July 1, 2026.
Why were Claude Fable 5 and Mythos 5 blocked in the first place? The Commerce Department issued an export control directive on June 12 citing national security concerns. Reports later indicated Amazon had flagged a jailbreak vulnerability in Fable 5 a technique involving asking the model to identify software vulnerabilities in code which reportedly prompted the government action.
What did Anthropic agree to in order to get the ban lifted? Anthropic agreed to proactively detect and address security risks in its models, work with the US government on protocols for future frontier model releases, and report any malicious activity it discovers.
Will this pattern of government sign-off repeat for future AI models? There’s no official policy confirming this, but the pattern is emerging both Anthropic’s Mythos 5 and OpenAI’s GPT-5.6 launched under government-coordinated rollouts, suggesting frontier AI releases are increasingly subject to US government oversight even without a formal regulatory framework in place yet.
Sources: Bloomberg, 9to5Mac, Al Jazeera, Engadget, Anthropic’s official X post, and Financial Times reporting, July 1, 2026.
