An executive signals a quick return for the Fable 5 and Mythos 5 models, which were pulled offline last week due to a national security directive linked to a corporate partnership program.
A sudden global shutdown of Anthropic’s most advanced AI models, Fable 5 and Mythos 5, could end within days, a top executive said. The outage, which began last week, was triggered by an emergency U.S. national security directive.
The abrupt suspension rattled the AI industry, highlighting the growing friction between tech innovation and government oversight. The press event, meant to celebrate Anthropic's South Korean expansion, was instead dominated by questions about the U.S. export controls.
The U.S. government’s security concerns appear to center on Mythos 5, a model with extraordinary, and potentially dangerous, capabilities.
Unlike consumer-facing AI, Mythos 5 can scan, identify, and repair severe vulnerabilities in highly complex, supposedly unbreakable code. This dual-use nature means it could also discover and exploit critical infrastructure flaws, making its access a significant national security issue.
To manage this risk, Anthropic restricted access to Mythos 5 through an exclusive corporate cybersecurity initiative called "Project Glasswing." The program includes around 150 partners, such as Google, NVIDIA, Microsoft, and Apple.
The high-stakes conflict leaves Anthropic's paying customers in limbo, awaiting a resolution. Ciauri deflected most questions on the topic, keeping the details of negotiations with Washington private.
The immediate challenge is whether Anthropic can implement new regional guardrails and security patches that satisfy the U.S. government. How the company resolves this standoff could set a critical precedent for the future of global AI partnerships and regulation.
