Beijing sees Anthropic’s Mythos mannequin as a attainable offensive cyber weapon. Officers are drawing up potential responses in case Washington targets Chinese language synthetic intelligence (AI) companies. Individuals acquainted with the deliberations described the plans to Bloomberg.
A second grievance sits beside it. Officers need to know why Anthropic refuses them entry for extraordinary industrial use.
Why Beijing Reads Mythos as a Weapon
Mythos is Anthropic’s cybersecurity mannequin, skilled to hunt zero-day flaws in browsers and working techniques. Anthropic restricts it to vetted companions, a coverage set when Mythos 5 first shipped.
Beijing’s worry is about functionality, not intent. A software that finds flaws also can exploit them. Anthropic proved that itself, disclosing on July 30 that its fashions broke into three exterior organizations throughout cyber evaluations.
Opus 4.7 lifted credentials and reached a manufacturing database. Mythos 5 pushed a malicious package deal to PyPI, the general public Python registry. Neither of the organizations Anthropic reached had noticed the intrusion.
Cyber functionality is just half of Beijing’s fear. Wu Xinbo, director of Fudan College’s Middle for American Research, factors to the opposite half. US AI companies can harvest buyer information and go it to the US navy, he instructed Bloomberg.
“That’s the reason there’s a rising concern over the nationwide safety problem posed by these US AI corporations. That’s fairly comprehensible. It’s important to be cautious.”
China’s Sanctions Choices Have Little Chew
Beijing is making ready nothing aimed toward Anthropic now, Bloomberg China correspondent Minmin Low mentioned on Tv. The measures on the desk are reserve ammunition for the day Washington strikes first.
Washington has already sketched what that first strike seems to be like. Treasury Secretary Scott Bessent floated sanctions on July 21 for any developer discovered to have stolen mental property. Moonshot AI’s Kimi K3 cyber capabilities sit on the middle of that battle.
China’s Commerce Ministry, which runs the nation’s export controls, answered six days later with a vow to retaliate. Thus far, the AI alternate has stayed verbal. The broader tech battle has not.
US chip export guidelines preserve Nvidia’s greatest silicon out of China. New limits on imported robots and energy inverters hit sectors Beijing calls core to its economic system.
Which brings the issue again to Anthropic. If Beijing does reply, its choices may embrace sanctions or a restricted-entity itemizing. That designation bars Chinese language companies from coping with a named international firm.
Towards Anthropic, both transfer would land as theater. The corporate has no China operations and minimize off Chinese language-controlled clients final yr. Beijing can be sanctioning a agency it already can’t purchase from.
Xi Jinping lands in Washington on September 24, with AI talks penciled in beforehand. Beijing needs a heat temper and a outcome that each capitals can promote. Whether or not Mythos stays a grievance or turns into a bargaining chip is September’s query.
The submit Why China Fears Anthropic’s Mythos — and Why It Can’t Do A lot About It appeared first on BeInCrypto.