US Lifts Export Controls on Claude Fable 5 - Model Returns to All Users on July 1

Anthropic announced on June 30, 2026, that the US government has lifted the export controls it had applied to Claude Fable 5 and Claude Mythos 5¹. Fable 5 returned to users worldwide the next day, July 1, across the Claude Platform, Claude.ai, Claude Code, and Claude Cowork¹. This brings an initial resolution to an unusual episode in which the company’s most capable model disappeared from public access for roughly three weeks after the controls took effect on June 12.

As part of the resolution, Anthropic deployed an improved safety classifier that blocks the jailbreak technique that triggered the restrictions in over 99% of cases¹. US Secretary of Commerce Howard Lutnick announced that Anthropic “has agreed to proactively detect and address security risks” and will collaborate with the government on protocols for these models². The lifting of the controls thus comes packaged with voluntary security commitments from the company. There is no precedent for the US government intervening this directly in a frontier AI model and then reversing course this quickly, making the episode an important reference point for how AI and regulation will interact.

Restricted Three Days After Launch - A Three-Week Timeline

Here is the sequence of events¹²:

Date (2026)	Event
June 9	Fable 5 and Mythos 5 released
June 12	US government applies export controls; Anthropic suspends public access
June 26	US government approves restoring Mythos 5 access for a set of US organizations
June 30	Export controls lifted; improved safety classifier deployed
July 1	Fable 5 becomes available again to users worldwide

Fable 5 and Mythos 5 are the Anthropic models announced on June 9. Their distinctive release approach split a single underlying model into two tracks: Fable 5, a generally available version with safeguards enabled, and Mythos 5, a restricted version with cybersecurity safeguards removed for vetted organizations.

Just three days after launch, on June 12, the US government placed both models under export controls¹. According to TechCrunch, the models were added to a list of export-restricted technologies, prohibiting access by foreign nationals without special approval². Enforcing nationality-based access on a service used worldwide proved “impractical at scale,” and Anthropic was forced to end public access entirely².

The trigger for the restrictions was a jailbreak discovered by Amazon researchers¹. The technique could get Fable 5 to identify software vulnerabilities, and in one case “the model produced code demonstrating how the relevant vulnerability could be exploited”¹.

The Terms of the Lifting - Voluntary Measures and Government Cooperation

Announcing the removal of the controls, Commerce Secretary Lutnick said Anthropic has agreed to proactively detect and address security risks and will work with the government on protocols for these models². Rather than mandatory reviews or permanent restrictions, the resolution rests on the company strengthening its own safeguards.

The centerpiece of Anthropic’s response is the improved safety classifier. It targets and blocks the behavior described in the report, shutting down the specific technique from the Amazon report in over 99% of cases¹. Anthropic also launched a new HackerOne program where security researchers can submit potential cyber jailbreaks they discover in Fable 5¹. The aim is to catch similar issues early through bug bounties and outside scrutiny rather than government restrictions.

Toward an Industry-Wide Jailbreak Severity Score

The response goes beyond one company’s fix. Anthropic said it will work with Amazon, Microsoft, Google, and other partners in Project Glasswing — the cyber-defense program run in cooperation with the US government — on an industry-wide framework for scoring the severity of jailbreaks¹.

The framework evaluates a discovered jailbreak on four criteria¹:

Capability gain: how far the technique pushes the model beyond existing tools
Breadth of capability gain: how many distinct offensive tasks it enables
Ease of weaponization: how much human effort an actual attack would require
Discoverability: how readily the technique can be found or reproduced

Since jailbreak reports are inevitable, a shared yardstick for “how severe does it have to be before availability is cut off” can be seen as a mechanism to avoid repeating this kind of abrupt, full-scale shutdown.

Availability - Up to 50% of Weekly Usage Limits Free Through July 7

Following the redeployment, Fable 5 is included in Pro, Max, Team, and select Enterprise plans for up to 50% of weekly usage limits through July 7, after which it will be available via usage credits¹. Re-enablement on cloud platforms — AWS, Google Cloud, and Microsoft Foundry — is being expedited¹.

Mythos 5 access was restored earlier for a set of US organizations following the government’s approval on June 26¹. On the same June 30, Anthropic also announced Claude Sonnet 5, a lower-priced model, and Claude Science, a research workbench — signaling both a return to normal operations and a renewed product push.

The Relationship Between Frontier AI and Government Remains Unsettled

The episode unfolded right after the AI security executive order signed on June 2. That order framed pre-release reviews of frontier models as strictly voluntary and explicitly ruled out mandatory licensing³ — yet ten days later, the government reached for export controls, a far blunter instrument. The inconsistency has not gone unnoticed: TechCrunch wrote that “the Trump administration’s erratic approach to AI policymaking has left companies across the industry with little clarity about what will govern future model releases”².

Some cybersecurity experts questioned whether the restrictions served as political leverage rather than genuine security measures². It has also been reported that Asian competitors began releasing comparable models (Fugu and Tulongfeng) during the restriction period, adding pressure to ease the controls in order to maintain American AI competitiveness².

For businesses that rely on AI, the episode leaves two lessons. First, there is now a precedent for a government decision abruptly cutting off access to a commercial AI model — a concrete reason to keep fallback options for workflows that depend heavily on a single model. Second, the resolution took the form of company-led safeguards plus an industry-wide evaluation framework. Whether a future incident ends in a graduated response instead of a full shutdown will depend largely on whether the new severity-scoring framework works in practice.

For details on Fable 5’s capabilities and the two-track release approach, see our coverage of the original announcement. With the free allocation of up to 50% of weekly usage limits running through July 7, this is a good window to try the top-end model.