Trump administration bars foreign access to Anthropic's Mythos and Fable models

The Trump administration moved last week to restrict foreign access to Anthropic's advanced Mythos and Fable AI models after concerns surfaced about potential security vulnerabilities. Amazon chief executive Andy Jassy contacted Treasury Secretary Scott Bessent to report that researchers had identified a way to bypass the safety measures in Fable, prompting officials to apply export controls that bar use by non-U.S. nationals. Anthropic responded by taking both models offline entirely rather than continue under the new limits.

Administration officials expressed frustration with the company's handling of the issue. They noted that Anthropic had received multiple warnings from other firms about risks tied to Fable before the controls were enacted. One senior official described the company's stance as inconsistent, pointing out its past public warnings about broad AI dangers while disputing the need for immediate restrictions in this case. Another official said Anthropic had chosen the wrong path at key decision points, including proceeding with deployment despite known issues.

The episode highlighted ongoing difficulties in communication between the company and government regulators. Sources familiar with the discussions said Anthropic struggled to align its approach with administration priorities on national security and technology oversight. The company maintained that it had secured prior approval for Fable's release and viewed the identified flaw as narrow rather than systemic.

Markets reacted quickly to the curbs. Shares of Chinese AI developer Zhipu rose sharply, climbing as much as 48 percent before settling around 33 percent higher. Analysts at JPMorgan and Bank of America cited the development as a potential opening for domestic Chinese firms, with Zhipu planning to release its latest GLM-5.2 model as open-source software without usage restrictions. The move underscores how limits on American companies can shift competitive advantages abroad.

Anthropic's premium offerings have drawn steady demand from individual and business users willing to pay several hundred dollars monthly for advanced access. Some subscribers report using the tools to support side projects and professional tasks, though the recent disruption leaves those arrangements in limbo. The episode illustrates the trade-offs involved when regulatory actions affect widely adopted technologies.

Critics within the administration argued that the company's emphasis on hypothetical long-term risks has not always matched its practical responses to immediate concerns. This pattern, they suggested, has complicated efforts to establish clear guidelines for high-capability systems. The controls remain in place as discussions continue over how to balance security requirements with continued development of the technology.