Anthropic announced late on Tuesday that the United States has removed export controls that had blocked foreign access to its most advanced models, Claude Fable 5 and Mythos 5. The company said it will begin restoring access the following day, following notification from the US Department of Commerce. The Commerce Department told Anthropic it would no longer require an export licence after the firm agreed to enhanced risk-mitigation steps and to cooperate with government standards. The move ends a temporary, high-profile restriction that had disrupted research and commercial access to the models.
Key takeaways
- On Tuesday, the US Department of Commerce lifted export restrictions on Anthropic’s Claude Fable 5 and Mythos 5, allowing foreign access to resume under new conditions.
- Anthropic said it will begin restoring access the day after the Commerce notification; the company thanked users and partners for patience during the pause.
- Commerce Secretary Howard Lutnick indicated no export licence would be required after Anthropic agreed to detect security risks, coordinate on standards and report malicious activity.
- Anthropic had shut the models last month after a White House directive ordered the company to block access by all foreign nationals; the company said officials cited unspecified national security concerns.
- The models were previously approved for limited use by US organisations that operate or defend critical infrastructure, per Anthropic’s earlier statements.
- Industry observers say the reversal reduces regulatory uncertainty for frontier labs but raises questions about how often the US government will demand pre-approval for major releases.
Background
The restriction on Anthropic’s largest models grew out of a government effort to scrutinise cutting-edge AI for risks to national security. In June, the White House ordered Anthropic to deny access to Fable 5 and Mythos 5 to all foreign nationals, prompting the company to disable those models for users outside the United States. Anthropic said at the time the administration offered no detailed justification beyond broad national security concerns, and the shutdown prompted immediate debate across industry and research communities.
Anthropic has positioned itself as a safety-conscious competitor in an industry racing to develop increasingly capable systems. Earlier this year the company negotiated limited approvals to provide the models to US organisations involved in critical infrastructure defence, while publicly pushing back against demands to integrate its technology into military workflows without strict safeguards. That stance contributed to friction with parts of the US government and provoked legal and regulatory sparring.
Main event
Late on Tuesday the company posted a short statement on its social channels saying access restoration would begin the next day after the Department of Commerce informed Anthropic it had removed its export controls. Anthropic thanked users and those who assisted in redeploying the models. The Commerce notification — circulated online — said Anthropic would not require an export licence after agreeing to new commitments around proactive detection of security risks, collaboration on standards for future models, and notification of malicious use.
The decision follows an earlier, narrower approval that allowed the company to serve the models to US organisations described as operating and defending critical infrastructure. Anthropic told partners it was working with the government to broaden access beyond that limited pool. The Commerce action effectively clears a path for researchers, companies and customers outside the United States to regain access under the terms the firm agreed to with regulators.
The disruption had occurred amid a broader tightening of US oversight of AI technologies. In parallel, other leading labs have taken cautious, staged approaches to new releases: OpenAI recently limited initial exposure of its GPT-5.6 series to a small set of trusted partners after similar government pressure. The Anthropic episode emerged as a focal point of industry concern about ad hoc export controls and the precedent they might set for future frontier-model rollouts.
Analysis & implications
The Commerce Department’s conditional lifting of controls signals a compromise: regulators can insist on operational safeguards without permanently barring foreign access to major models. For Anthropic, the agreement preserves market access while obliging the firm to strengthen monitoring and engagement with government scrutiny. That balance will likely be tested by future model updates and by any demonstrated misuse of capabilities.
For the wider AI sector, the episode underscores that Washington is prepared to intervene aggressively at points it views as risky. The reversal, however, suggests the government recognised costs to overly broad or indefinite blocks — including diplomatic friction, research setbacks, and industry backlash. Companies developing frontier systems will now face clearer pressure to demonstrate technical risk controls and to participate in regulatory dialogues before large releases.
Internationally, the decision reduces the chance of a fragmented landscape where certain frontier models are wholly inaccessible outside the United States. Yet it also raises questions about parity: if US models are subject to conditional access, other states may pursue reciprocal restrictions or demand their own safeguards. The net effect could be a more complex, multi-jurisdictional permissioning system for high-capability AI.
Comparison & data
| Model | Prior status | Current status | Access conditions |
|---|---|---|---|
| Claude Fable 5 | Blocked to foreign nationals | Access being restored | Anthropic to detect risks, report malicious activity |
| Mythos 5 | Blocked to foreign nationals | Access being restored | Same conditions as Fable 5 |
| GPT-5.6 | Phased release to trusted partners | Limited rollout | Staggered access after government pressure |
The table summarises the immediate statuses after the Commerce decision. While Fable 5 and Mythos 5 move from a broad blockade to conditional availability, competitor releases like GPT-5.6 remain subject to staged partner rollouts. These differences reflect each company’s regulatory engagement and internal risk-management choices; the practical impact will depend on the technical controls actually implemented and on enforcement of reporting commitments.
Reactions & quotes
Industry analysts welcomed the restoration as preferable to an open-ended export ban, noting it reduces uncertainty that can chill investment and international research collaborations. At the same time, experts emphasised that the efficacy of the resolution will depend on the technical and operational measures Anthropic implements and on ongoing transparency with regulators.
We’re grateful to our users for their patience, and to everyone who worked with us on redeploying the models.
Anthropic statement (social post)
Anthropic framed the development as a cooperative step to restore services. The company’s short public message highlighted its focus on resuming access while acknowledging the work that went into meeting the government’s conditions.
We will not require an export licence now that Anthropic has agreed to proactively detect and address security risks and to work with us on standards.
Howard Lutnick, US Department of Commerce (letter summary)
The Commerce letter summarised the basis for removing the licence requirement, tying the decision to explicit commitments by Anthropic. Officials characterised those commitments as operational safeguards intended to reduce the risk of misuse.
Unconfirmed
- Whether specific technical vulnerabilities in Fable 5 were identified by the Commerce Department has not been publicly disclosed.
- Reports that researchers had widely successful “jailbreaks” of Fable 5 appear to be exaggerated; their precise scope and impact are not independently verified.
- Details of the monitoring and reporting mechanisms Anthropic will deploy under the agreement have not been released in full.
Bottom line
The Commerce Department’s decision to lift export controls on Claude Fable 5 and Mythos 5, conditional on Anthropic’s commitments, resolves a disruptive episode that had threatened to curtail international access to two leading models. It establishes a template in which regulators can demand operational safeguards rather than choosing permanent exclusion, but the durability of that template depends on implementation and enforcement.
For industry and policymakers alike, the episode highlights the need for clearer, pre-established processes for assessing frontier models so that decisions are predictable and proportionate. Observers should watch closely how Anthropic fulfils its obligations and whether the administration seeks to apply the same approach to future frontier-model releases.