The U.S. government ordered Anthropic to immediately disable access to two of its most advanced AI models — Claude Fable 5 and Claude Mythos 5 — on Friday, citing national security concerns. Anthropic confirmed compliance in a statement on X, but argued publicly that the government had misjudged the situation.
According to Anthropic, the directive arrived at 5:21 PM ET and requires the company to block both models for all users globally, not just foreign nationals targeted by the original export control order. Other Anthropic models remain unaffected.
Why This Matters
Mythos is Anthropic’s most capable AI system to date. First previewed in early April 2026, the company kept it tightly restricted due to what it described as Mythos’s exceptional ability to identify security vulnerabilities in software. Anthropic reported that Mythos successfully found flaws in every major operating system and web browser it tested. Rather than release it broadly, the company launched a controlled initiative called Project Glasswing, sharing the model with approximately 50 vetted organizations, including Amazon, Apple, Google, Microsoft, and CrowdStrike — all for defensive cybersecurity use.
Fable 5, released just three days earlier, was Anthropic’s response to commercial demand: a version of Mythos fitted with guardrails that block responses in high-risk domains like cybersecurity and biology. The company argued this made it safe for general release. Benchmark tests from Vals AI, which tracks AI performance, immediately ranked Fable 5 as the most capable AI model available to the public.
The Irony of Safety Warnings
Anthropic’s own safety warnings may have contributed to the government’s decision. By publicly emphasizing Mythos’s ability to find critical security flaws, the company may have inadvertently alarmed regulators. In a post on X, Anthropic stated: “Our transparency about Mythos’s capabilities was meant to build trust and inform responsible use, not to trigger a shutdown that limits access for everyone.”
Industry and Policy Implications
This development has major implications for AI governance in 2026. The U.S. government’s decision to halt access to state-of-the-art AI models — even those with built-in safety measures — signals a more aggressive stance on AI regulation. Critics argue this could stifle innovation and push companies to become less transparent about their models’ capabilities. Supporters, however, see it as a necessary precaution to prevent the proliferation of tools that could be exploited for cyberattacks.
The shutdown affects organizations worldwide that relied on Fable 5 and Mythos 5. Anthropic is reportedly in discussions with the government to restore access, but no timeline has been provided. The company emphasized that it remains committed to safety and collaboration with regulators, even as it disagrees with this specific decision.
via TechCrunch AI