Alibaba ‘bans’ Anthropic in AI Cold War as cybersecurity regulation heats up in China
Alibaba ‘bans’ Anthropic in AI Cold War as cybersecurity regulation heats up in China
Publish Date: 2026-07-06 07:28:00
Source Domain: www.sdxcentral.com
Using an unordered list, summarize the following article with between 4 and 8 key points. Chinese tech giant Alibaba reportedly banned its workforce from using Anthropic large language models (LLMs) as an AI Cold War escalates between the U.S. and China.According to Reuters, Alibaba outlawed Anthropic’s Claude model due to an alleged backdoor in its underlying code. The escalation comes after Anthropic recently accused Alibaba of illicitly extracting Claude capabilities, claiming it had carried out almost 29 million exchanges with its frontier model via fraudulent accounts.As of Friday, Alibaba employees have reportedly been told to stick to the firm’s Qoder coding platform. While Reuters did not have particulars of the order, which was originally reported by Chinese publication Yicai, the development came after developers outed alleged spyware in Claude Code, which checks a user’s timezone and proxy-related data for possible Chinese intrusion. The claims were made in a Reddit post that the obfuscated code examines whether the system timezone matches China’s, or looks up a proxy URL against a list of Chinese domains and Chinese AI labs it may be associated with.Anthropic confirmed the feature was an experimental one launched in March to block unauthorized access and distribution, and defend against model distillation: the process of distilling knowledge from a superior AI model into a lower-capacity one, of which it accused Alibaba of doing this month.The AI Cold WarReuters’ source who verified the Yicai report added that the ban would be difficult to enforce if Alibaba employees chose to deploy U.S. servers on the sly to run Claude models.This would put Alibaba in the crosshairs of Chinese regulation. While China may seem like a Wild West of deregulation, the country recently revamped the China Cybersecurity Law (CSL) to better protect its own sovereignty on the digital frontier.That revamp may have led to Alibaba’s Claude ban; Western companies like Nokia operating in the region have been taking it seriously enough to note their full compliance online with the law.The CSL recently saw China initiate a “Qinglang” or clean-up campaign targeting unregulated and unsafe AI models and usage, which could have conceivably included unauthorized distribution of foreign LLMs.Away from legislation, Chinese cybersecurity players have also been revving up development to take Anthropic head-on. This month saw local vendor Qihoo 360 unveil what it called “China’s version of Mythos” – the advanced Claude model, which Anthropic claimed has a supernatural skill for finding vulnerabilities that human security experts have overlooked.Anthropic has not been immune to the China question, with a version of its Mythos model alongside Fable 5 ending up banned from export by the U.S. government due to a jailbreak vulnerability, with potential repercussions for national security should it fall into enemy nations.As of last week, the ban was fully lifted on Fable 5, while Mythos 5 was partly freed up to be available for an exclusive group of U.S. organizations.
