The company says a federal order limiting access to advanced AI technology forced it to pull its newest models on Friday, but the government has not stated what national security concern drove the action, and Anthropic is disputing the process.
Anthropic pulled its newest AI models offline on Friday in response to a federal directive aimed at preventing their use by foreign nationals, but the company says the government did not name the national-security concern that drove the order, and Anthropic is publicly disputing the way the decision was reached.
The takedown, first reported by SecurityWeek, covers two systems SecurityWeek identifies as Fable 5 and Mythos 5 — a naming convention Anthropic has confirmed on its own announcements page as of June 12, 2026. Fable 5 was released widely the same week; Mythos 5 is a more restricted variant Anthropic had already been limiting access to over cybersecurity concerns.
Anthropic told SecurityWeek it received the directive Friday afternoon and complied the same day. The company argues that any block of unsafe deployments should happen through a "transparent, fair, clear, and grounded in technical facts" statutory process, and characterized the action as a "misunderstanding." The full company statement, as available in the captured source, is truncated around that characterization, so the precise scope of Anthropic's complaint has not been verified.
The federal actor behind the directive has not been confirmed against a primary government source. SecurityWeek describes the action as coming from the Trump administration and as the most significant step the U.S. government has taken to date to restrict access to the most advanced AI models. Which administration body issued the order, and under what statutory authority, remains to be verified. No direct readout from the Commerce Department's Bureau of Industry and Security, the Commerce Department itself, or the White House was part of the source basis available for this draft.
That gap is the part of the story a wire summary tends to miss. A regulator compelling a frontier AI developer to take its newest models offline on a Friday afternoon is the kind of action that sets the working pattern for any future takedown. It also lands without a stated national-security rationale, on a public record that does not yet specify which statutory authority the directive is operating under.
There is a structural tension in the public record worth noting. President Donald Trump signed an executive order on June 2, 2026 establishing a framework for federal vetting of the most advanced frontier AI models before their public release. That framework, as SecurityWeek reported, is explicitly voluntary — participation by AI developers is optional, and Section 3(c) of the order states nothing in it shall be construed to authorize mandatory governmental licensing, preclearance, or permitting requirements for the development, publication, release, or distribution of new AI models. The directive Anthropic received, by contrast, was compelled: the company took its models offline the same day it received the order. Whether this action was taken under a different legal authority than the June 2 executive order — or whether the voluntary framework has effectively become mandatory in practice — is a question the public record does not yet answer.
The practical effect, for now, is concrete. Two of Anthropic's newest models are no longer available to users, including researchers and developers outside the United States. The company has not contested the underlying authority to restrict access to advanced AI systems. It has contested the process, and the public record on which that process rests.