Anthropic Denies Chinese Think Tank Access to Its Newest AI Model, Claude Mythos

Anthropic rebuffed an attempt from a Chinese think tank to gain access to its newest AI model, Claude Mythos, during a private meeting in Singapore. The New York Times reported the rebuff on May 12, 2026. Anthropic unveiled Mythos in April 2026 but has limited release of the model to around 40 American companies.

Those companies include Amazon, Apple and Microsoft. The company acted over concerns that Mythos is so powerful it could threaten national security systems and potentially disable companies that use it. The request came during a meeting organized by the Carnegie Endowment for International Peace in April 2026.

A representative from the Chinese think tank asked Anthropic officials to loosen their stance and give Beijing access to the Claude Mythos AI model. The request for access to Claude Mythos was not an official ask from the Chinese government. Trump administration officials at the National Security Council were alarmed by the Chinese think tank's request for access to Claude Mythos.

Some officials saw the request as a sign that Beijing will try to infiltrate American companies by any means possible to gain access to new and powerful AI models. China is scrambling to build out an AI model as powerful as Claude that is focused on cyber defense. 7 billion) by 2030.

During a meeting with German lawmakers earlier in May 2026, Claudia Plattner warned that Chinese tech companies were beginning to go dark to develop a model like Mythos that has “superhacking” abilities. Claudia Plattner is Germany’s top cybersecurity official. 5-Cyber.

In January 2026, the Department of War said in a memo that it would only contract with AI companies that agreed to “any lawful use” and would be willing to remove safeguards involving surveillance and the development of autonomous weapons. Anthropic refused to accept the Department of War contract terms. U.S.