In Beijing, concerns over cybersecurity are growing, and Anthropic's Mythos model has come under the spotlight. Chinese officials view this AI tool as a potential cyber weapon and are already preparing countermeasures in case Washington imposes new restrictions on Chinese AI companies. This situation is not just another round of technological rivalry, but a signal that the AI race is escalating to the level of military-strategic confrontation.

Why Mythos Raises Alarm

Mythos is a model designed to find zero-day vulnerabilities in browsers and operating systems, access to which Anthropic provides only to a limited circle of trusted partners. A tool capable of finding breaches, by definition, can also exploit them. Beijing's concerns are not unfounded: in late July, Anthropic confirmed that during cyber tests, its own models penetrated three third-party organizations, and one of them did not even notice the breach. Opus 4.7 extracted login credentials and reached the working database, while Mythos 5 uploaded a malicious package to the public PyPI registry.

The second issue raised by Wu Xinbo, director of the Center for American Studies at Fudan University, is the risk of American AI companies collecting client data and transferring it to the U.S. Department of Defense. "Anxiety is growing due to the threat to national security. This is quite understandable. We need to be cautious," the expert notes.

Countermeasures: For Now, Only Threats

Despite the rhetoric, Beijing is not yet ready for active measures. The measures under discussion are more of a reserve in case Washington makes the first move. In the U.S., proposals have already emerged to impose sanctions on developers caught stealing intellectual property, and at the center of the confrontation are the cyber capabilities of Moonshot AI's Kimi K3. China's Ministry of Commerce, responsible for export controls, has promised to take countermeasures, but so far it has all been limited to statements.

Notably, for Anthropic, sanctions would be more of a symbolic gesture: the company does not operate in China and already stopped servicing clients controlled by the PRC last year. In essence, Beijing would impose restrictions on an organization from which it already buys nothing. However, against the backdrop of Xi Jinping's visit to Washington on September 24, where AI negotiations are planned in advance, Beijing seeks to maintain an appearance of friendliness and present any agreements as a success.

My analysis: This situation is a vivid example of how AI tools are becoming a new arena for geopolitical struggle. For the crypto industry, where code and data security are critically important, the escalation of such conflicts could lead to stricter regulations and fragmentation of technological ecosystems. Investors should closely monitor developments: any sanctions measures in the AI sphere will inevitably affect adjacent markets, including blockchain projects that depend on American cloud services and AI infrastructure.