Anthropic suspends Fable 5, Mythos 5 access under US export control order
Synopsis
The US government has ordered Anthropic to pull its most advanced AI models — Fable 5 and Mythos 5 — from all foreign nationals globally, citing a jailbreak risk. Anthropic has complied but is pushing back hard, warning the standard used could freeze every frontier AI deployment in America. It is one of the most consequential direct government interventions in commercial AI to date.
Key Takeaways
Anthropic suspended global access to Fable 5 and Mythos 5 on 13 June following a US government export control directive .
The order bars all foreign nationals — including Anthropic employees — from accessing the models, inside or outside the United States .
The government cited a potential 'jailbreak' of Fable 5 as the national security basis, but did not share specific details.
Anthropic reviewed the technique and found it exposed only 'minor, previously known vulnerabilities' replicable by other publicly available models.
Pre-launch red-teaming with the US government and UK AISI found no universal jailbreak across thousands of hours of testing.
Anthropic warned the government's standard, if applied industry-wide, would effectively halt all new frontier model deployments in the US.
Anthropic has confirmed that the US government issued an export control directive ordering the company to immediately suspend access to its advanced AI models — Fable 5 and Mythos 5 — for all foreign nationals, including its own employees, whether based inside or outside the United States. To comply, Anthropic has abruptly disabled both models for all customers globally, the company disclosed in an official blog post on 13 June.
What the Government Ordered
The directive, issued under US export control authority, bars any foreign national from accessing the two frontier models. Anthropic said the government did not furnish specific details of the underlying national security concern, but indicated that authorities believe a method of bypassing — or 'jailbreaking' — Fable 5 has been identified. The company was given no prior opportunity to remediate the issue before the suspension order took effect.
Anthropic's Rebuttal
Anthropic publicly disagreed with the government's conclusion, stating that 'the finding of a narrow potential jailbreak should be cause for recalling a commercial model' deployed to millions of users is not a standard the company considers proportionate. The company reviewed a demonstration of the specific technique and found it surfaced 'a small number of previously known, minor vulnerabilities.' Critically, Anthropic noted that 'other publicly-available models are able to discover them as well without requiring a bypass,' undercutting the argument that the vulnerability is unique to its systems.
Months of Pre-Launch Safety Testing
In the weeks before Fable 5's launch, Anthropic worked with the US government, the UK AI Safety Institute (AISI), multiple private third-party organisations, and internal red teams to stress-test the model's safeguards for thousands of hours. Those evaluations, the company said, showed that Fable 5's safety guardrails are 'substantially more effective than those of any previously deployed model.' Notably, no tester across those exercises was able to find a universal jailbreak — one capable of broadly bypassing the model's safeguards and unlocking a wide range of cyber capabilities.
Industry-Wide Implications
Anthropic warned that if the government's standard — recalling a commercial model on the basis of a narrow jailbreak finding — were applied uniformly across the industry, it would 'essentially halt all new model deployments for all frontier model providers.' This signals a potentially significant chilling effect on AI development at a moment when US companies are racing to maintain a competitive edge over international rivals. The episode marks one of the most direct government interventions in frontier AI deployment to date, raising urgent questions about the regulatory framework governing advanced AI models and the balance between national security and commercial continuity.
What Happens Next
Anthropic has not indicated a timeline for restoring access to Fable 5 and Mythos 5. The company's public pushback suggests it may seek to challenge or negotiate the directive. Millions of users and enterprise customers who rely on these models face an indefinite disruption, and the broader AI industry will be watching closely to see whether other frontier labs receive similar orders.
Point of View
Disputed jailbreak finding. Anthropic's public rebuttal is unusually direct for a company that has cultivated a cooperative relationship with regulators. The deeper tension here is structural: export control law was designed for hardware and code, not for API-served inference, and applying it to model access creates a legal grey zone that neither the government nor the industry has fully mapped. If this directive stands unchallenged, it sets a precedent that any jailbreak demonstration — however minor — could trigger an immediate global recall, effectively handing a veto over AI deployment to national security agencies with no transparency requirement. That is a governance design no one has publicly agreed to.
NationPress
29 Jul 2026
Frequently Asked Questions
Why did the US government order Anthropic to suspend Fable 5 and Mythos 5?
The US government issued an export control directive citing a potential method of bypassing, or 'jailbreaking,' Fable 5 as a national security concern. The government did not provide specific details of the threat, but ordered Anthropic to immediately suspend access for all foreign nationals globally.
What did Anthropic find when it reviewed the jailbreak technique?
Anthropic reviewed a demonstration of the specific technique and found it revealed only a small number of previously known, minor vulnerabilities. The company noted that other publicly available AI models can discover the same vulnerabilities without requiring any bypass, suggesting the risk is not unique to Fable 5.
Who is affected by the suspension of Fable 5 and Mythos 5?
All customers globally have had access to both models abruptly disabled. The directive specifically targets foreign nationals — including Anthropic's own employees — inside and outside the United States, but Anthropic disabled the models for all users to comply.
What safety testing did Anthropic conduct before launching Fable 5?
In the weeks before launch, Anthropic red-teamed Fable 5 with the US government, the UK AI Safety Institute, multiple private third-party organisations, and internal teams for thousands of hours. No tester found a universal jailbreak capable of broadly bypassing the model's safeguards.
What are the broader implications of this directive for the AI industry?
Anthropic warned that if the government's standard — recalling a model based on a narrow jailbreak finding — were applied across the industry, it would effectively halt all new frontier model deployments in the United States. The case is being closely watched as a potential precedent for government intervention in commercial AI.
