Anthropic, the AI safety company behind the Claude model, has turned down a request from a Chinese think tank for access to its AI systems. The rejection, which came without a public explanation from either side, underscores the tightening environment around cross-border sharing of advanced artificial intelligence.
The Quiet Refusal
The think tank, a research organization based in China, made the request directly to Anthropic. Details of the models sought or the intended use have not been disclosed. The company's decision was confirmed by people with knowledge of the matter, but neither Anthropic nor the think tank has issued a statement.
Why Access Was Denied
Anthropic has long maintained policies limiting access to its most capable AI models. The company cites safety and security risks, and often restricts use by entities in countries with export control restrictions on dual-use technologies. This rejection fits that pattern. The U.S. government has increasingly scrutinized AI transfers to China, citing potential military or surveillance applications. While the think tank's work is presumed to be civilian, the decision reflects a broader caution among frontier AI labs.
A Growing Trend in AI Labs
Anthropic is not alone. Other major AI developers, including OpenAI and Google DeepMind, have also blocked access requests from Chinese organizations in recent years. The moves come amid rising geopolitical tensions and a push by Washington to keep cutting-edge AI out of Beijing's hands. For Chinese researchers, this means fewer opportunities to work with the latest models—a gap that may widen as export controls tighten.
What Comes Next
The think tank has not indicated whether it will appeal the decision or seek access through other channels. Anthropic has not said if it plans to review its access policies more broadly. For now, the rejection stands as a quiet but firm signal: even academic requests from China face hurdles in the current climate. No timeline for any further action has been announced.
