Anthropic Restores Fable 5 and Mythos 5 as US Lifts Controls

Anthropic turned its two most advanced AI models back on for the public after the US government lifted the export controls that forced their removal in June. Fable 5 began rolling out again worldwide on July 1, 2026.

The company confirmed the reversal in a post on X and a newsroom notice, saying the controls on both Claude Fable 5 and Claude Mythos 5 were removed on June 30. CNBC reported that Commerce Secretary Howard Lutnick said an export license would no longer be required for the models.

What the US Government Reversed

The Commerce Department dropped export controls it had imposed roughly 18 days earlier. CIO quoted Anthropic's blog post stating that as of June 30 the controls on Fable 5 and Mythos 5 had been lifted.

Fable 5, the public-facing model, is returning globally across the Claude Platform, Claude.ai, Claude Code and Claude Cowork. Access on Amazon Web Services, Google Cloud and Microsoft Foundry will follow in phases.

Mythos 5 is a separate case. CoinDesk reported that the more capable, less restricted variant is being restored only to approved US organizations after a government review that cleared it on June 26.

Why the Models Went Dark

The freeze dated to June 12, when the government applied export controls to both models. The rules limit which foreign nationals can access a technology, and because the order took effect immediately, Anthropic suspended access for everyone rather than risk a breach.

The trigger was a cybersecurity finding. As Technobezz reported, Amazon researchers surfaced a jailbreak that pushed Fable 5 to identify software vulnerabilities and, in one instance, produce code showing how one could be exploited.

Anthropic disputed the severity throughout. In its notice to users the company said the demonstration involved a small number of previously known, minor vulnerabilities that other publicly available models could also surface, which it framed as a borderline case for Fable 5's safeguards.

The distinction between the two models is why the alarm spread. Mythos 5 is the powerful cybersecurity system, and Fable 5 is the consumer-facing layer built on top of it, so a bypass of one raised questions about reaching the capabilities of the other.

The Timeline in Brief

June 9 was the launch, when Anthropic released Fable 5 for general users and Mythos 5 for a limited group of Project Glasswing cybersecurity partners. Both share the same underlying model, with Fable 5 shipping the stronger guardrails.

June 12 brought the shutdown. MediaNama reported that the government imposed the controls after the Amazon report, and Anthropic pulled both models within hours.

June 26 was partial restoration, when Washington approved bringing Mythos 5 back for select US organizations. June 30 closed the loop as the controls were lifted, clearing Fable 5's July 1 return.

The Fixes Anthropic Shipped

Anthropic did not simply flip a switch. The company trained a new safety classifier that it says blocks the specific bypass technique in more than 99 percent of cases, while acknowledging the tighter protections will raise false positives for some legitimate coding requests.

Researchers at the Commerce Department's Center for AI Standards and Innovation evaluated the updated safeguards. In exchange for the reversal, Anthropic agreed to proactively detect security risks, help develop standards for future models, and report malicious activity to the government, according to The Next Web.

The company is also proposing a shared method for scoring how dangerous a jailbreak is, developed with Amazon, Microsoft, Google and other Glasswing partners. It plans a round-the-clock monitoring team and a HackerOne program for researchers to report cyber jailbreaks in Fable 5.

The rollout itself is deliberately gradual. Anthropic said Fable 5 would be capped at up to 50 percent of normal weekly usage limits through July 7 before returning to full availability, and Pro, Max, Team and select enterprise users receive extra weekly usage as compensation for the disruption.

What It Means for People Building on Frontier AI

The episode shows that model access can be shaped by policy as much as by capability. A control meant for foreign nationals became a worldwide outage because nationality could not be verified in real time, which is a new kind of operational risk for buyers.

The dependency runs deeper than one vendor. As the debate over rules intensifies, our coverage of who gets to police AI in America shows how quickly the ground can shift for companies that assumed a stable regulatory backdrop.

Security is the other half of the story. The jailbreak that started this sits inside a wider reckoning that we examined in our reporting on AI transformation and security, where model safety and enterprise trust increasingly move together.

For teams that standardize on Claude for coding and agent work, the practical takeaway is to keep a tested alternative ready and to plan for stricter classifiers. Restored access is not restored certainty, and the road now runs through policy.