Anthropic export ban forces shutdown of Mythos and Fable 5 as US imposes controls
U.S. export controls targeting Anthropic’s Mythos and Fable 5 have prompted the company to shut both models and block access for all users, citing national security concerns and a reported jailbreak vulnerability. The Anthropic export ban—announced by the U.S. government—prohibits foreign nationals from using the two most powerful models, a move that the company says includes foreign employees located in the United States. In response, Anthropic took the unusual step of disabling access for its entire customer base and suspending the models while the situation is reviewed.
U.S. government orders export controls on Mythos and Fable 5
The administration issued export restrictions that, according to Anthropic, bar non-U.S. persons from accessing Mythos and Fable 5, even if they are physically present in the United States. The scope of the measures is unusually broad for an AI export action and extends beyond typical technology-transfer limits to cover usage by individuals. Anthropic said the company received limited detail from officials but was told the move was driven by national security concerns related to the models’ capabilities.
Anthropic shuts models and blocks access worldwide
Faced with the order and uncertainty over compliance, Anthropic opted to block all customers from using Mythos and Fable 5 and took both systems offline temporarily. The company’s decision to disable access for U.S. customers as well reflects the complexity of enforcing nationality-based restrictions on cloud-delivered services. For now, both models remain shuttered while Anthropic and regulators assess remediation steps and legal options.
Security research flagged advanced offensive capabilities
Independent testing by the U.K. AI Security Institute found that Mythos demonstrated markedly stronger cybersecurity capabilities than prior models, including the ability to autonomously complete a simulated multi-step attack against a corporate network. Those findings, and similar tests showing parity between Mythos and other leading models, intensified scrutiny by governments and private-sector researchers. Anthropic had already limited Mythos access to a small set of U.S. companies for testing because of misuse risks tied to the model’s ability to discover and exploit software vulnerabilities.
Reports of a ‘jailbreak’ prompted official alarm
Anthropic told officials there was concern about a method that could bypass Fable 5’s safety filters, a so-called jailbreak that would let the model provide information useful for cyberattacks. Media reports have linked the discovery of that vulnerability to researchers at Amazon and say company executives raised the issue with senior government officials. While details remain opaque, the alleged exploit appears to have been a central factor cited by regulators in imposing the export restrictions.
Anthropic questions selective enforcement and faces Pentagon dispute
The company has publicly questioned why controls singled out its models when similar vulnerabilities can be found across other leading systems. Anthropic pointed to tests indicating comparable performance by models from other firms and noted the lack of parallel measures against those competitors. The dispute takes place against a backdrop of a months-long confrontation between Anthropic and the U.S. Defense Department; late this winter the Pentagon labeled the company a national security concern, and Anthropic subsequently sued the government, deepening tensions.
Short-term business impact and implications for an IPO
Because Mythos and Fable 5 were available only to a small set of pilot customers, the immediate commercial damage to Anthropic may be limited. However, the timing is sensitive: the company has been preparing for a public offering and is under pressure to demonstrate rapid growth and scalable products. Anthropic’s projections for annualized revenue have recently risen in public statements, and any prolonged shutdown of flagship models could complicate investor perceptions and market timing.
European industry reactions and push for AI sovereignty
The export action has catalyzed concern in Europe about dependence on U.S. AI providers. Industry groups and policymakers warned that access to advanced models can hinge on foreign government decisions, with potential consequences for industrial competitiveness, public-sector services, and research. European leaders and trade associations have called for accelerated investment in native AI capabilities, with initiatives under discussion that include public-private funding pledges and coordinated research consortia to develop continental foundation models.
Emerging responses: projects, funding and open-source strategy
European plans under consideration include collaborative model development projects hosted on national cloud infrastructure and proposals to mobilize large-scale investment to bolster local capabilities. Officials are weighing open-source approaches that could reduce lock-in to closed U.S. systems by making model architectures, weights and training code more transparent. At the same time, some Chinese providers signaled they would seek to fill gaps in global availability, while the EU said it would evaluate the broader impact of the export measures on European users and industries.
For now, the Anthropic export ban underscores a new reality in which geopolitical and security considerations can directly reshape commercial AI availability and strategy. Governments, companies and researchers will watch closely as regulators clarify the grounds for the restrictions, Anthropic pursues legal and technical responses, and European stakeholders press for a faster path to technological autonomy.
