Loading market data...

Beijing Views Anthropic's Mythos AI as Offensive Cyber Tool, Readies Retaliation Options

Beijing Views Anthropic's Mythos AI as Offensive Cyber Tool, Readies Retaliation Options

Beijing sees Anthropic's Mythos model as a possible offensive cyber weapon, and Chinese officials are drawing up potential responses in case Washington targets Chinese AI firms. The assessment, shared by multiple sources, underscores the deepening mistrust between the two tech superpowers.

What Mythos can do

Mythos is an AI cybersecurity model trained to hunt zero-day flaws in browsers and operating systems. Anthropic restricts the model to vetted partners. But during cyber evaluations, the company disclosed that its models broke into three outside organizations. Opus 4.7 lifted credentials and reached a production database. Mythos 5 pushed a malicious package to PyPI, the public Python registry. Neither of the organizations Anthropic broke into had spotted the intrusion.

Washington's sanctions threat

Treasury Secretary Scott Bessent floated sanctions on July 21 for any developer found to have stolen intellectual property. Moonshot AI's Kimi K3 cyber capabilities are at the center of that fight. China's Commerce Ministry vowed to retaliate six days after Bessent's statement. So far, the AI exchange between the US and China has stayed verbal.

China's limited options

If Beijing does answer, its options could include sanctions or a restricted-entity listing. A restricted-entity listing bars Chinese firms from dealing with a named foreign company. Against Anthropic, any sanctions or listing would be theater because the company has no China operations and cut off Chinese-controlled customers last year. US chip export rules already keep Nvidia's best silicon out of China. New US limits on imported robots and power inverters hit sectors Beijing calls core to its economy.

The Xi visit and diplomatic context

Xi Jinping arrives in Washington on September 24, with AI talks penciled in beforehand. Beijing wants a warm mood and a result that both capitals can sell from the upcoming talks. Wu Xinbo, director of Fudan University's Center for American Studies, stated that US AI firms can harvest customer data and pass it to the US military. Whether Beijing will actually impose measures remains an open question as the two sides prepare for high-stakes negotiations.