Anthropic is restoring access to its Claude Mythos and Fable 5 AI models after the Commerce Department lifted export controls.
The restrictions began last month when the agency directed Anthropic to disable access for foreign nationals. The move followed the Pentagon’s decision to label the company a supply chain risk, along with broader concerns about its advanced models. This designation placed Anthropic under heightened scrutiny, as the government sought to mitigate potential risks associated with the proliferation of its AI capabilities to non-U.S. entities.
Related: Memory Chip Makers Face Price Fixing Lawsuit
Mythos Preview had already drawn attention in April after the company reported it could identify vulnerabilities across every major operating system and web browser, including some dating back 27 years. The model’s ability to uncover long-dormant flaws in widely used software demonstrated its potential to expose critical weaknesses in global digital infrastructure. In June, CEO Dario Amodei acknowledged the model’s risks, calling it a potential threat to cybersecurity, financial systems, and national infrastructure. His remarks showed the growing recognition within the industry that frontier AI models could have disruptive implications far beyond their intended applications.
Weeks later, the government ordered Anthropic to block foreign nationals—even its own employees—from using Mythos and Fable 5. The company complied but noted the directive lacked a clear justification.
Related: Walmart’s 4K Streaming Box Gets Major Upgrade With Gemini and Matter Support
Access restoration began today, with further updates expected as the company works to reintegrate users.
In a follow-up blog post, Anthropic emphasized its safeguards against jailbreaking attempts, arguing that bypassing the model’s safety measures remains difficult. The company highlighted its investment in robustness testing and adversarial evaluations to prevent malicious actors from circumventing built-in protections. However, it did not address why the government initially imposed the restrictions or what triggered their removal, leaving the underlying rationale for the controls unresolved.
Related: A Beginner’s Guide to Software: Unlocking the Digital World
The Pentagon’s supply chain risk designation and the Commerce Department’s sudden reversal highlight the evolving, often opaque oversight of advanced AI systems. The shifting regulatory stance reflects the tension between supporting innovation and mitigating potential harms, as policymakers grapple with the dual-use nature of cutting-edge AI. Anthropic’s models, capable of uncovering long-standing software flaws, have become a focal point in this debate, demonstrating both the transformative and destabilizing possibilities of AI development.
As access resumes, the lack of official explanation for the initial controls leaves lingering questions about how such decisions are made—and how long they might last. The episode shows the need for clearer communication between regulators and AI developers to avoid future disruptions and ensure a more predictable framework for managing emerging technologies.