China Considers Restricting AI Model Access

Chinese authorities are considering whether to restrict foreign access to advanced artificial intelligence models developed by domestic companies like Alibaba, ByteDance, and Z.ai. According to Time, officials have held discussions with these firms about potentially limiting the public release of their most sophisticated AI models or restricting them for domestic use only.

The potential restrictions mark a significant shift from China's current practice of releasing AI models 'open-weight,' allowing anyone to download and run the systems. This openness has helped Chinese companies gain global market share despite trailing U.S. models by about seven months on key benchmarks, as reported by Time. The models are typically free, with some U.S. businesses adopting them to cut costs compared to proprietary American alternatives.

The discussions come amid escalating tensions between China and the United States over AI technology. According to CNBC, Alibaba has banned employees from using Anthropic's AI tools for work purposes, citing concerns about back-door security risks. This move follows Anthropic's accusation that Alibaba attempted to extract its AI capabilities through what it termed 'the largest known distillation attack' on the company.

Meanwhile, Ars Technica reports that Anthropic recently removed a tracker from Claude Code after a security researcher exposed hidden code monitoring Chinese users. The tracking mechanism was designed to detect potential connections to Chinese AI labs and prevent unauthorized use of the technology. This incident has sparked controversy over user trust and privacy.

In an updated warning, CNBC reports that China's Ministry of Industry and Information Technology identified a security back-door vulnerability in Claude Code versions 2.1.91 to 2.1.196. The ministry advised users to uninstall or upgrade from these affected versions, which were released between April 2 and June 29.

According to CBS News, a Chinese industry regulator warned users of a 'security backdoor' embedded in versions of U.S. artificial intelligence giant Anthropic's coding tool, Claude Code. The alleged backdoor could enable the software to transmit sensitive information, including users' locations and identity-related identifiers, back to Anthropic's servers without users' consent. The National Vulnerability Database (NVDB), affiliated with China's Ministry of Industry and Information Technology, advised relevant institutions and users to conduct a comprehensive check immediately and promptly uninstall or upgrade to the latest secure version from which the relevant backdoor code has been removed.