The United States has moved to sharply restrict access to some of the world’s most advanced artificial intelligence systems, triggering a sweeping suspension by Anthropic of its flagship models Fable 5 and Mythos 5. The decision follows an export control directive issued by the Trump Administration requiring the company to block foreign nationals from accessing its newest AI tools, citing national security and cybersecurity concerns.
Anthropic said it received formal notice on Friday at 5:21 p.m. Eastern Time from Commerce Secretary Howard Lutnick instructing it to prevent distribution of its most advanced models to foreign nationals, including individuals residing inside the United States. The directive also extends to foreign employees working at the company, who are now unable to access systems they helped develop. The order effectively created a compliance challenge that Anthropic said it could not resolve through targeted restrictions, prompting the company to suspend broad public access to the models altogether.
The move marks one of the most far-reaching export control actions yet applied to artificial intelligence technology in the United States. It reflects a growing approach within Washington that treats frontier AI systems as strategic assets with potential national security implications. Officials have cited the increasing sophistication of AI tools and their ability to identify vulnerabilities in computer systems as a central concern, warning that such capabilities could be exploited for cyberattacks or intelligence gathering.
Fable 5 and Mythos 5 sit at the center of that debate. Mythos 5, described as a highly advanced model with the ability to detect vulnerabilities in complex systems, had already been placed under strict limitations. Its derivative, Fable 5, was introduced as a more controlled version intended for broader public use through Anthropic’s Claude service. The company had positioned it as a “safe” model designed for general tasks such as software design and scientific research, while incorporating safeguards intended to prevent misuse.
The suspension comes only weeks after Fable 5 was made available to subscribers, a rollout that Anthropic itself described as a carefully constrained expansion of access. The company had also previously established Project Glasswing to restrict use of Mythos 5 to a limited number of organizations in critical sectors, including finance, energy, and telecommunications, as well as select government partners. Those efforts were aimed at identifying vulnerabilities and preventing security breaches while still allowing controlled experimentation with the system’s capabilities.
However, the US government’s latest directive appears to have concluded that even limited exposure to such systems presents unacceptable risks. According to Anthropic, the government has suggested it identified a potential security vulnerability in Fable 5, though the company says it has not been given detailed explanations for the national security concerns behind the decision.
The order has reignited debate over how AI systems should be regulated and who should be allowed access to their most powerful versions. Export controls have become a key instrument in Washington’s strategy to manage emerging technologies, particularly as AI systems are increasingly integrated into sensitive domains. Government officials have described advanced AI as both a technological breakthrough and a potential cybersecurity threat, capable of accelerating both defensive and offensive cyber operations.
Anthropic, led by CEO Dario Amodei, has pushed back against the decision, arguing that the government has not clearly justified the withdrawal of its models. The company said it disagrees with the premise that the discovery of a limited vulnerability should be sufficient grounds for removing a product used by hundreds of millions of people. It also warned that applying such a standard across the industry could significantly slow or halt the release of new frontier AI systems.
Tensions between Anthropic and the US administration have been building for months. The Department of Defense previously labeled the company a “supply chain risk,” a designation that restricted its ability to secure federal contracts. That dispute followed Anthropic’s refusal to allow unrestricted military use of its AI systems, particularly in relation to mass surveillance and autonomous weapons without human oversight. While earlier conflicts eventually eased, the latest export controls suggest a renewed escalation in regulatory scrutiny.
The impact of the decision has already extended beyond the United States. The European Commission has acknowledged the suspension, stating it is reviewing the implications of the US directive while continuing discussions with technology partners regarding cybersecurity risks associated with advanced AI models. European institutions had been working with Anthropic to explore potential access to its systems for use in industries such as banking and infrastructure.
At the heart of the controversy are concerns about the dual-use nature of frontier AI. Anthropic has acknowledged that models like Fable 5 possess significant cybersecurity capabilities, including the ability to identify weaknesses in complex computer systems. The company has also admitted that, under certain conditions, such capabilities could be misused if not properly controlled, particularly in high-risk sectors such as financial services.
Despite these risks, Anthropic maintains that it has implemented strict safeguards designed to limit misuse. It has also stated that many users have criticized these protections as overly restrictive. In internal testing, the company says no universal method has been found to bypass its security systems at scale, although it concedes that the models are still under active evaluation for potential vulnerabilities.
When Fable 5 was launched, Anthropic itself acknowledged that misuse could have serious consequences if the model were deployed without restrictions. In response, certain outputs were deliberately limited or redirected to less capable systems to reduce potential harm. The company described this approach as part of its broader safety framework for managing frontier AI tools.
Even as the suspension takes effect, neither the pace of AI development nor the regulatory scrutiny surrounding it appears to be slowing. The decision leaves Fable 5 and Mythos 5 inaccessible to the public and foreign users for the foreseeable future, marking a significant moment in the tightening global controls over advanced artificial intelligence systems.
