Britain Urged to Reject Cheap Chinese AI Over Data Risks and Censorship
Updated
Updated · The Guardian · Oct 17
Britain Urged to Reject Cheap Chinese AI Over Data Risks and Censorship
3 articles · Updated · The Guardian · Oct 17
Summary
British businesses and public services are being urged to avoid low-cost Chinese AI models, with critics warning they could embed Beijing’s political controls into systems used for information and decision-making.
DeepSeek’s downloaded models still showed political censorship when run on local infrastructure, according to a US NIST evaluation cited by the Centre for Information Resilience, undermining the argument that self-hosting removes the main risk.
DeepSeek’s chatbot policy says prompts, uploaded files and personal data can be processed and stored in China and disclosed to Chinese authorities; Oxford professor Michael Wooldridge said it is reasonable to assume user inputs could reach the Chinese state.
The warning goes beyond privacy: advocates say subsidized Chinese AI could undercut UK and European suppliers just as Chinese EVs have pressured foreign industries, leaving Britain more dependent on Chinese technology.
If self-hosting cheap AI doesn't stop data leaks or built-in censorship, are businesses unknowingly installing foreign surveillance directly onto their private servers?
While the West worries about data theft, are heavily subsidized AI models actually a Trojan horse designed to rewrite global digital infrastructure?
With strict 2026 laws banning these models for defense contractors, could your company's cost-saving AI strategy secretly disqualify you from lucrative government contracts?