Anthropic abruptly disabled access to its Fable 5 and Mythos 5 AI models on Friday evening after receiving a U.S. Commerce Department directive, citing national security concerns and reports of a potential jailbreak bypassing safeguards.
Anthropic’s move—announced in a late-night post—marks the first time the company has halted a model launch after receiving government orders, and comes just days after the models were made public. The shutdown affects all users, including foreign nationals and Anthropic employees, leaving other models like Claude untouched. The Commerce Department’s directive, received at 5:21 p.m. ET, did not specify the exact security threat but referenced concerns from “national security authorities” about a jailbreak that could expose vulnerabilities in critical software.
What the Government Ordered—and Why
According to Ars Technica, the Commerce Department’s directive stemmed from reports of a “narrow, non-universal jailbreak” in Fable 5—one that could bypass safeguards designed to block prompts related to cybersecurity, chemistry, and biology. The administration reportedly sought a pause to “harden” its national security apparatus against such threats, with officials suggesting the process could take “weeks.” Anthropic’s statement confirmed the government provided only “verbal evidence” of the jailbreak, which the company described as involving “minor” and “relatively simple” software vulnerabilities—similar to capabilities seen in other publicly available models like GPT-5.5.
Yet the directive’s timing and breadth stand out. The models were launched just days prior, with Fable 5 touted as Anthropic’s first state-of-the-art public offering, built on safeguards that had already impressed Wall Street and government officials. The abrupt shutdown—affecting even foreign employees—suggests the government’s concerns outweighed the company’s assurances about controlled access. As CNBC noted, the order arrived at 5:21 p.m. ET, leaving Anthropic little time to negotiate or appeal before disabling access entirely.
How This Fits Into Anthropic’s Larger Clash With the Government
This isn’t the first time Anthropic has clashed with the U.S. government. Earlier this year, the Department of Defense labeled the company a “supply chain risk” to national security—a designation historically reserved for foreign adversaries. The move forced defense contractors to certify they wouldn’t use Anthropic’s models in military work, and the company responded by suing the Trump administration to reverse the blacklisting. Litigation is ongoing, but the latest shutdown suggests the government’s concerns have only deepened.
Anthropic’s public stance on the matter underscores the tension: “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,” the company stated. “This action does not adhere to those principles.” The phrasing hints at frustration—not just with the shutdown, but with the lack of clarity around the government’s decision. Without specific details on the jailbreak or the risks it poses, Anthropic’s ability to defend its models—or even restore them—remains uncertain.
The Broader Implications: Who Wins, Who Loses?
For Anthropic, the shutdown is a setback but not a total loss. The company’s other models, including Claude, remain operational, and its cybersecurity-focused initiative, Project Glasswing, continues with limited access. Yet the move raises questions about the government’s ability to preemptively block AI deployments—and whether such actions could stifle innovation. As Ars Technica pointed out, the government’s reliance on “verbal evidence” of a jailbreak sets a precedent for future restrictions, potentially allowing officials to halt models based on incomplete or unverified claims.
For competitors like OpenAI, the shutdown could be seen as a cautionary tale—or an opportunity. If the government continues to tighten controls on advanced AI, companies may face greater scrutiny over safeguards, model capabilities, and even employee access. Meanwhile, users relying on Fable 5 and Mythos 5 for specialized tasks—such as cybersecurity research or scientific modeling—are left without immediate alternatives. The shutdown also complicates Anthropic’s plans for 2026, particularly if the government’s concerns persist or expand to other models.
What Happens Next: The Timeline and Uncertainties
The next few weeks will be critical. The government’s “hardening” process—whatever that entails—could lead to a partial or full restoration of Fable 5 and Mythos 5, or it could result in permanent restrictions. Anthropic’s lawsuit against the DOD adds another layer of uncertainty, as the outcome could influence how future disputes between tech companies and regulators are resolved. For now, the company has apologized to customers for the disruption but offered no timeline for resolution.
One thing is clear: the shutdown signals a turning point in AI governance. The government’s ability to act swiftly—and without full transparency—sets a precedent that could reshape how companies develop and deploy advanced models. For Anthropic, the challenge will be balancing compliance with innovation, while for users, the question remains: how long will they have to wait for access to be restored?
In the meantime, the incident serves as a reminder of the high stakes in AI development. With national security concerns growing, the line between progress and risk has never been more blurred.
Find more reporting in our Business section.
