Project Glasswing

Anthropic is signaling that its most advanced AI models could return soon after an abrupt freeze triggered by an emergency U.S. national security directive. The pause has drawn attention because it affects systems considered powerful enough for high-risk cybersecurity work. The company said it is “very confident” the models will be available again in the coming days, according to remarks by Chris Ciauri, Managing Director of International Anthropic, at a press event in Seoul. That statement offers the clearest indication yet that the disruption may be temporary rather than permanent. The freeze went far beyond one market. Anthropic pulled the Fable 5 and Mythos 5 models globally, leaving paying customers without access to the company’s top-tier systems while it works through the security concerns. Why Mythos 5 drew U.S. concern Mythos 5 sits at the center of the dispute because it is designed to scan, identify, and repair severe vulnerabilities in very complex code bases. That makes it valuable for defensive cybersecurity, but also potentially useful for finding and exploiting weaknesses in critical infrastructure if misused. To limit that risk, Anthropic restricted access through Project Glasswing, its exclusive corporate cybersecurity initiative. The program serves as the main gatekeeper for who can use the model. Anthropic says Project Glasswing has around 150 partners, including Google, NVIDIA, Microsoft, and Apple. The list underscores how widely the company has tried to place controls around access to its most capable systems. How the standoff with Washington escalated Wired reported that U.S. export controls hardened after a dispute with Anthropic over SK Telecom’s entry into Project Glasswing. The South Korean telecom operator was said to have gained access to Claude Mythos 5. According to people familiar with the matter, U.S. officials were worried about alleged ties between SK Telecom and China. That concern was reportedly a key factor behind the emergency freeze on Anthropic’s advanced models. Anthropic declined to comment on the Wired report. The White House and SK Telecom also did not respond, although a SK Telecom representative told The Chosun Daily that unnamed claims in foreign media were not verified and that the company has no ties to China. What happens next for customers For paying customers, the immediate issue is access. The company’s most capable models are unavailable for now, and the next step appears to depend on whether Anthropic’s regional safeguards and security patches are enough to ease Washington’s concerns. No official technical details have been published about those additional protections. Still, the direction of the discussions suggests that any restoration will likely depend on a combination of geographic controls and new security layers. The episode shows how advanced AI models are increasingly treated as strategic assets rather than ordinary commercial products. As AI becomes more deeply tied to cybersecurity, the line between innovation, technology exports, and geopolitics is becoming much thinner. Anthropic is also in the middle of international expansion, which makes the Seoul remarks especially significant. The company appears eager to reassure partners and customers that the disruption will not last long, even as attention remains fixed on when Fable 5 and Mythos 5 will actually come back online. If the models do return within days, as Anthropic expects, the decision will become an early test of whether its new safeguards are enough to satisfy U.S. national security concerns. Source: www.androidauthority.com

Anthropic’s Most Advanced AI Is Frozen, but a Return May Be Days Away