Anthropic disabled access to its two most advanced AI models, Fable 5 and Mythos 5, on Friday evening after receiving an export control directive from the U.S. government citing “national security authorities.”
The company said it received the order at 5:21 p.m. ET on Friday, instructing it to suspend all access to the models “by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees,” according to Anthropic’s statement. To ensure compliance, Anthropic disabled the models for its entire customer base, not just foreign users. All other Anthropic models remain accessible, the company said.
The Three Facts That Matter
- The order arrived days after launch. Anthropic had announced Fable 5 and Mythos 5 just days before the shutdown, touting them as state-of-the-art across multiple industry benchmarks. Fable 5 was notable as the first time Anthropic had released such an advanced model to the general public, enabled by new safeguards designed to block outputs in specific high-risk areas. The models built on the April release of Claude Mythos Preview, which Anthropic said drew significant interest from Wall Street and government officials because of its advanced cybersecurity capabilities. For context on the underlying model architecture, see our earlier explainer on Anthropic’s Claude Fable 5 and why it has two versions.
- The government offered no specific justification. Anthropic said in its statement that the government did not provide specific details about its national security concern. The company apologized to customers for the disruption while simultaneously pushing back on the process: “As we have stated publicly, we believe the government should have the ability to block unsafe deployments, as part of a statutory process that is transparent, fair, clear, and grounded in technical facts. This action does not adhere to those principles,” Anthropic said. The statement represents a rare instance of a leading AI company simultaneously complying with a government directive and publicly contesting its legitimacy in the same breath.
- This is Anthropic’s second major collision with U.S. authorities in 2026. The shutdown is the latest episode in a deteriorating relationship between Anthropic and the U.S. government. Earlier this year, the Department of Defense declared Anthropic a supply chain risk — a designation historically reserved for foreign adversaries — following the collapse of negotiations between the two organizations. That label requires defense contractors to certify that they will not use Anthropic’s Claude models in their work with the military. Anthropic subsequently sued the Trump administration to reverse the blacklisting, and that litigation remains ongoing, according to the company. The Pentagon dispute had already drawn attention to competitive dynamics between OpenAI and Anthropic following the Pentagon fallout.
Taken together, Friday’s shutdown and the ongoing DOD blacklist litigation reveal a structural tension that no leading AI company has yet resolved: the more capable a frontier model becomes — particularly in domains like cybersecurity — the more likely it is to attract national security scrutiny, even from the government the company is attempting to partner with. Anthropic’s Mythos Preview had been specifically praised for its cybersecurity capabilities and was already being rolled out under a restricted initiative called Project Glasswing; those same capabilities appear to be precisely what made the broader Fable 5 and Mythos 5 release a regulatory flashpoint. The cycle suggests that capability improvements at the frontier may now reliably generate legal and regulatory friction, not just commercial opportunity. This dynamic is worth watching alongside Anthropic CEO Dario Amodei’s broader public stances on AI governance, which have grown increasingly pointed in recent months.
Anthropic’s case stands out among major AI companies because it is the first major U.S. AI developer to reportedly face both a government-directed suspension of flagship models and a supply chain risk designation. While other companies such as OpenAI and Google DeepMind have faced export controls and government scrutiny, they have largely responded by complying with restrictions and building closer partnerships with federal agencies. Anthropic, however, challenged the process legally while complying with the directive, creating uncertainty over whether its more confrontational approach will strengthen or hurt its position as AI regulation and infrastructure competition intensify.
How Serious Players Should Respond
For enterprise customers and developers who built workflows on Fable 5 or Mythos 5, the immediate priority is operational: identify fallback models, assess contract exposure, and monitor Anthropic’s legal filings for timeline signals. Anthropic has confirmed all other models remain available, but the sudden nature of this suspension is a vivid reminder that dependency on a single frontier provider carries sovereign risk that standard business continuity planning rarely accounts for.
For institutional investors, the episode adds a new risk dimension to Anthropic’s pre-IPO narrative. Both Anthropic and OpenAI have been reported as racing toward public markets; a pattern of government conflict — including active litigation against a presidential administration — is the kind of regulatory overhang that underwriters and institutional allocators will need to price carefully. The DOD supply chain risk label, in particular, is not a routine compliance matter; it is the same classification framework used against foreign adversaries.
For policymakers and regulators, Anthropic’s public statement deserves careful attention. The company complied with the order while explicitly stating it did not meet the company’s own standards for transparency, fairness, or technical grounding. That posture invites a broader legislative conversation about whether the United States has the right statutory framework for managing frontier AI capabilities — one that can balance genuine national security imperatives against the procedural legitimacy that even cooperative companies are now demanding in writing.
