Anthropic has decided to withdraw a covert monitoring feature designed to track whether users were located within China’s borders or had ties to Chinese artificial intelligence laboratories, following a wave of criticism on social media.<\/p>\n
Although the US-based AI company restricts access to its services in China, Claude models are known to be widely used by developers in the country.<\/p>\n
Earlier this week, a post on Reddit alleged that Claude Code, Anthropic’s software development tool, was secretly collecting and transmitting information indicating whether a user was located in China and whether they were affiliated with a Chinese AI laboratory.<\/p>\n
An analysis published by cybersecurity newsletter International Cyber Digest also reported that Claude Code embedded data such as users’ time zones, proxy usage and potential links to AI laboratories into system prompts using methods that Chinese users would not easily detect.<\/p>\n
The issue also gained traction on social media platform X, where numerous Chinese-speaking users criticized Anthropic over the practice.<\/p>\n
Thariq Shihipar, a member of Anthropic’s technical team working on the Claude Code project, confirmed the practice in a response to International Cyber Digest’s post.<\/p>\n
Shihipar said the feature was “an experiment launched in March to prevent account abuse by unauthorized resellers and to protect against model distillation.”<\/p>\n
He added that the team had already been planning to remove the functionality and that it would be phased out completely in the near future.<\/p>\n
Anthropic has previously accused Chinese AI companies DeepSeek, Moonshot AI, MiniMax and Alibaba Group of illegally using responses generated by Claude models to train their own large language models and of conducting large-scale model distillation attacks.<\/p>\n
Although the company does not officially offer its models and services in China for national security reasons, many developers in the country continue to access Claude by using overseas phone numbers and credit cards.<\/p>\n
<\/p>\n","protected":false},"excerpt":{"rendered":"
Anthropic has decided to withdraw a covert monitoring feature designed to track whether users were located within China’s borders or had ties to Chinese artificial intelligence laboratories, following a wave of criticism on social media. Although the US-based AI company restricts access to its services in China, Claude models are known to be widely used […]<\/p>\n","protected":false},"author":1,"featured_media":15546,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"ep_exclude_from_search":false,"footnotes":""},"categories":[176],"tags":[19949,19947,9970,2122,529,19943,15965,18161,5442,19948,19945,6192,16719,19769,19946,19951,4900,19944,19952,19950,4901],"class_list":["post-15545","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-america","tag-account-abuse","tag-alibaba-group","tag-anthropic","tag-artificial-intelligence","tag-china","tag-chinese-users","tag-claude","tag-claude-code","tag-deepseek","tag-distillation-attacks","tag-international-cyber-digest","tag-large-language-models","tag-minimax","tag-model-distillation","tag-moonshot-ai","tag-proxy-usage","tag-reddit","tag-thariq-shihipar","tag-time-zone-data","tag-unauthorized-resellers","tag-x"],"yoast_head":"\n