Anthropic PBC has suspended global access to its advanced "Mythos" and "Fable 5" artificial intelligence models following a directive from the U.S. Department of Commerce. The government order, issued Friday, mandates that the company block all foreign nationals from accessing the systems due to national security concerns regarding potential cybersecurity vulnerabilities.
Why the U.S. Government Restricted Access
The Commerce Department intervened after identifying that the Fable 5 model could be "jailbroken"—a process where users bypass built-in safety guardrails to perform unauthorized tasks. According to a report by the Wall Street Journal, researchers at Amazon.com Inc. first identified these vulnerabilities and coordinated with federal officials before the restrictions were imposed.
Amazon CEO Andy Jassy reportedly engaged directly with senior U.S. officials regarding the risks. While an Amazon spokesperson confirmed the company frequently consults with the government on security, they declined to discuss specific exchanges.
Anthropic’s Response to the Directive
Anthropic, which positions itself as a leader in AI safety, officially disagreed with the government’s decision to pull the models from public use. In a statement posted to its website, the company argued that a "narrow potential jailbreak" does not justify a total recall of a commercial product deployed to millions of users.
"If this standard was applied across the industry, we believe it would essentially halt all new model deployments for all frontier model providers," the company stated. Anthropic received the order at 5:21 p.m. ET on Friday, forcing an immediate, total shutdown of access for both domestic and international users to ensure compliance.
Regulatory Precedents and Industry Impact
This action marks a shift in U.S. technology policy. While previous administrations have restricted the export of hardware like semiconductors and supercomputers, this represents a rare instance of the government targeting software itself.
The move creates significant uncertainty for major developers including OpenAI, Google, and Meta. David Sacks, a member of the President’s Council of Advisers on Science and Technology, stated on X that the administration’s goal is for Anthropic to remediate the safety issues so the export controls can be lifted and the models returned to general release.
Comparison of AI Policy Approaches
The current restrictions stand in contrast to the administration’s earlier, more collaborative approach to AI development.
| Policy Initiative | Focus | Status |
|---|---|---|
| Executive Order (Recent) | Voluntary model review | Remains in effect |
| Friday Directive | Mandatory access suspension | Currently enforced |
| Military/Defense Policy | Restricted use in surveillance | Ongoing tensions |
Earlier this year, the Pentagon designated Anthropic as a "supply-chain risk" following disagreements over the military application of its technology. Despite these tensions, the company had previously worked toward a limited release of Mythos for federal agencies, a move now stalled by the new security mandate.
What Happens Next
The European Commission has signaled it is assessing the impact of the U.S. order on its own technological sovereignty. Meanwhile, industry figures have expressed concern that the move could damage international cooperation. Aidan Gomez, co-founder of the Nvidia-backed startup Cohere, described the restriction as a "massive wake-up call" for the industry, noting that such unilateral actions may jeopardize long-standing technological alliances.
Anthropic continues to advocate for a standardized system where developers and governments coordinate on "pausing" potentially dangerous AI work, rather than relying on reactive, end-of-day mandates.