Anthropic has suspended access to two of its most advanced AI models after the U.S. government issued an extraordinary national security order restricting their use by foreign nationals, a move that thrusts frontier AI into the center of export control and security debates in Washington and beyond.
Lead and nut graf
Anthropic has disabled access to its Fable 5 and Mythos 5 models for all customers after the U.S. government issued an export control directive ordering the company to suspend access for any foreign national, citing national security authorities.
The order, delivered late Friday by the Trump administration, marks the first time Washington has forced a leading AI lab to pull a publicly available frontier model over security concerns, highlighting how advanced systems are being treated increasingly like controlled dual‑use technologies and sending a chilling signal through the global AI industry.
What the government ordered
In a statement on its website, Anthropic said the U.S. government had issued “an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.”
Because the company cannot technically separate domestic and foreign users in the short term, it said “the net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance,” even as access to all other Anthropic models remains unaffected.
Officials told media outlets that the letter came from the Commerce Department’s Bureau of Industry and Security, which oversees export controls on sensitive technologies.
Bloomberg reported that a U.S. official confirmed Commerce had instructed Anthropic to treat Fable 5 and Mythos 5 like controlled exports and to keep them out of the hands of all foreign nationals “whether inside or outside the United States.”
Anthropic’s response and “narrow jailbreak” dispute
Anthropic said it received the directive at 5:21 p.m. Eastern time and moved within hours to cut off Fable 5 and Mythos 5 globally, describing the order as a sweeping step tied to an unspecified national security concern.
The company said its understanding was that U.S. authorities believe they have identified “a narrow potential jailbreak” that could allow users to bypass some safeguards in Fable 5, potentially enabling assistance with hacking or other prohibited activities.
News reports indicate that external red‑teaming and research, including work by cloud partners, raised alarms that Mythos 5 and its derivative Fable 5 could be manipulated to generate sensitive cyber‑security output despite guardrails.
Anthropic has pushed back on the scope of the response, arguing that treating a targeted vulnerability as grounds to effectively freeze a whole model sets a precedent that could halt new deployments across the industry if applied consistently.
A widening clash with the Trump administration
The suspension order deepens a months‑long confrontation between Anthropic and the Trump administration over the company’s role in government and defense work.
In February, Trump directed all federal agencies to “immediately cease all use” of Anthropic’s technology, while the Pentagon labeled the company a “supply‑chain risk to national security,” a status historically used in actions against foreign telecom and surveillance firms.
Anthropic is challenging that earlier ban in court, and a judge has allowed some government use of its tools to continue while the case proceeds, but the new export control directive signals that the administration is prepared to escalate on multiple fronts.
SAN’s Straight Arrow News and other outlets reported that Commerce Secretary Howard Lutnick sent a letter to Anthropic CEO Dario Amodei declaring Fable 5 and Mythos 5 subject to export restrictions to any location outside the U.S. and to all foreigners within the country.
What changes for users and customers
Anthropic stressed that only Fable 5 and Mythos 5 are affected and that access to its other models, including widely used Claude systems, continues as normal.
Nonetheless, customers building products, research pipelines or internal tools on top of the two frontier models have seen those systems abruptly shut off, forcing rapid fallbacks to earlier models or rival providers.
The company said it has instructed cloud partners such as Amazon Web Services to revoke access to the two models for “all users in all regions” while it works with regulators to clarify compliance paths.
Anthropic has told customers it hopes to restore availability “as soon as possible,” but has given no public timeline, and an official quoted by Axios suggested the models could remain locked for weeks while U.S. agencies refine their security posture.
Export controls meet frontier AI
Policy experts say the directive is an early indication of how Cold War‑era export control tools are being stretched to cover AI models with potential offensive cyber, intelligence or military applications.
By treating access to Fable 5 and Mythos 5 by foreign nationals as an export, the U.S. is effectively placing advanced AI models in the same regulatory category as high‑end semiconductors, encryption software and other dual‑use technologies.
The Trump administration has already tightened controls on AI‑related chips and is weighing additional rules on data centers and training compute, but compelling a leading AI firm to switch off live models is a more visible intervention.
Supporters argue that such measures are necessary to prevent adversaries from using powerful models to discover software vulnerabilities, automate cyberattacks or guide weapons systems, while critics warn they could chill innovation and accelerate the development of non‑U.S. alternatives.
Safety branding versus coercive regulation
Anthropic has positioned itself as a “safety‑first” lab, investing in extensive red‑teaming, alignment research and access controls and marketing its Claude line as a more cautious alternative in the AI arms race.
The company has said Fable 5 and Mythos 5 were built with layered safeguards to reduce misuse, and that it had already implemented restrictions around security‑sensitive capabilities before the U.S. stepped in.
The episode exposes a tension at the core of AI governance: voluntary corporate safety practices may not be enough to reassure security agencies, which can reach for blunt tools like export bans when they perceive novel threats.
For Anthropic and its peers, the risk is that regulators demand more intrusive controls or pre‑deployment approvals for future generations of models, reshaping development timelines and access models for global users.
Global and US implications
For a U.S. audience, the Anthropic order crystallizes a domestic debate over how far national security concerns should reach into the commercial tech stack.
The decision could influence how agencies, companies and voters view Trump’s broader approach to technology regulation, balancing promises of innovation with fears about cyber threats and strategic competition with rivals such as China.
Globally, the move may reinforce concerns among allies and competitors that Washington is prepared to restrict access to cutting‑edge American AI even for friendly researchers and firms, encouraging efforts to build local or open‑source alternatives beyond U.S. jurisdiction.
Other governments could treat the Anthropic case as a template, crafting their own export controls or security regimes for advanced models and turning AI into another fault line in a fragmented digital economy.
As Anthropic negotiates with regulators and customers adapt to the sudden outage, the suspension of Fable 5 and Mythos 5 stands as an early inflection point in the politics of frontier AI—when the promise of general‑purpose models collided with the hard boundaries of national security law, and the U.S. government made clear it is willing to pull the plug.
