What growing tensions have emerged between China and the United States over Anthropic's advanced AI model Mythos, including China's fears it could be used as an offensive cyber weapon, U.S. export controls imposed in June after a suspected China linked breach, China's push for ac

Controls Lifted, but Tensions Remain

Less than three weeks later, on June 30, the Trump administration lifted the export controls after what Commerce Secretary Howard Lutnick described as a close collaboration with Anthropic to align the models with U.S. government approval . Critics at the Peterson Institute argued the ad hoc controls were counterproductive, potentially boosting China's motivation to develop indigenous alternatives .

China's Push for Access Ahead of the September Summit

China is pushing for access to Mythos — or at least assurances about its non-deployment — ahead of the planned summit between Xi Jinping and President Donald Trump in September 2026 . Beijing is growing increasingly anxious that the model could be used offensively against the world's second-biggest economy, adding a volatile new issue to already heightened bilateral tensions .

Broader AI Arms Race: Distillation, Open-Weight Models, and the Contest for Control

• Shrinking gap: U.S. estimates suggest America has only 6–12 months before China either gains access to a comparable frontier model (via distillation or theft) or develops its own competitor capable of being wielded as a cyber weapon .
• China's indigenous response: On June 25, 2026, 360 Security Technology claimed it had already developed a domestic AI system matching Mythos's capabilities, framing it as a strategic deterrent China could not afford to lack .
• Distillation concerns: The U.S. fears that China could use model distillation — extracting a frontier model's capabilities through repeated queries — or stolen weights to replicate Mythos without authorization, a concern that directly motivated the export clampdown .
• Open-weight debate: The episode has intensified global debate over whether the most powerful AI models should be open-weight and widely accessible, or tightly controlled as dual-use technologies akin to nuclear materials.

The Mythos standoff illustrates a fundamental shift: AI models that can autonomously find and exploit software vulnerabilities are now treated as strategic military assets, not just commercial products, and control over them is emerging as the defining technology contest between the U.S. and China.