Key Takeaways
- Chinese "open-weight" AI models (e.g., DeepSeek, Xiaomi) offer low cost but embed significant risks, including unavoidable political censorship and potential data sharing with Chinese state authorities.
- Hosting these models locally does not remove inherent censorship or security weaknesses, as confirmed by U.S. NIST evaluations finding political restrictions persisting in self-deployed versions.
- Relying on cheap Chinese AI risks hollowing out Western domestic tech industries, undermining data security, and advancing Beijing’s strategic goal of creating technology-dependent client states.
- Experts urge Western policymakers to reject the lure of low-cost Chinese AI, prioritize securing domestic and allied supply chains (US, EU), and implement stricter scrutiny on data handling and model origins.
The relentless march of Chinese artificial intelligence development presents a quiet but profound challenge to Western economic and security interests, disguised as an irresistible bargain. Every month, firms like DeepSeek and Xiaomi release new iterations of their "open-weight" AI models – software whose core code can be freely downloaded, modified, and deployed without ongoing licensing fees. This stands in stark contrast to the costly, proprietary systems offered by U.S. leaders like OpenAI and Anthropic. In Britain, where consumers have long embraced affordable Chinese manufactured goods – from electronics to increasingly, electric vehicles – the appeal of this "free" AI is potent. Critics dismiss concerns, asking what harm could come from accessing powerful tools at a fraction of the price? Yet this perspective overlooks the strategic calculus driving Beijing’s AI push, transforming a seemingly benign commercial offering into a potential vector for influence and vulnerability.
The affordability of Chinese AI is not merely a market advantage; it is deeply intertwined with the Chinese Communist Party’s evolving strategy for maintaining control and projecting power. A decade ago, internal weaknesses – a property market crash, rising unemployment, regional healthcare disparities, gender imbalances, and an aging population – threatened social stability and posed existential risks to Party rule, echoing the turmoil of Tiananmen Square. However, the state discovered a powerful new tool in artificial intelligence. AI now underpins vast surveillance capabilities, enabling the monitoring of citizen activities through image recognition and data analysis. While ordinary citizens retain freedoms to gather publicly for dance or local discussion, the system is ruthlessly efficient at identifying and neutralizing serious dissidents through re-education, sidelining, or punishment. For Beijing’s leadership, AI provides a mechanism to bypass the traditional financial constraints of appeasing a domestic populace, allowing them to focus resources on countering perceived external threats without the constant pressure of internal unrest.
This technological edge fuels a broader geopolitical ambition that extends far beyond internal security. China’s reinvigorated Belt and Road Initiative seeks not just to build infrastructure, but to establish enduring trade channels that turn partner nations into clients dependent on Chinese technology for basic economic and governmental functions. Unlike Russia’s overt use of AI to foment chaos via cyber-attacks on European targets, China’s approach is more insidious: it aims to widen the circle of states and businesses that routinely consume its goods and services, including AI, without probing the strings attached or questioning the underlying political imperatives. Concurrently, the resources freed by reducing domestic consumer subsidies are being redirected towards military modernization. Beijing is actively constructing a naval fleet designed to challenge U.S. dominance in the Pacific, echoing Japan’s pre-World War II ambitions, with Taiwan – the global epicenter of advanced semiconductor manufacturing – squarely in its sights as a critical prize for sustaining its AI and tech ambitions.
The tactical execution of this strategy is evident in the flood of low-cost AI models emerging from Chinese labs. While some analysts argue these models diversify the global AI landscape, supplementing closed U.S. systems and noting parallel efforts in Europe and India, the reality for potential Western users is stark. Choosing between a freely tweakable Chinese open system and expensive U.S. software laden with unnecessary features presents a dangerous false economy. Although both types of AI carry existential risks highlighted by experts, only the Western alternatives typically incorporate meaningful efforts towards security safeguards and benign operational intent – efforts conspicuously absent in the Chinese offerings. As Ross Burley, director of the Centre for Information Resilience (CIR), an independent non-profit focused on exposing human rights abuses and disinformation, bluntly stated: "Chinese AI comes with extra baggage." He elaborated that while running a model like DeepSeek on one’s own servers avoids direct access to the developer’s infrastructure, "that addresses just part of the risk. Running a model on your own servers doesn’t automatically remove any political censorship. Or indeed repair weaknesses in any security defences."
This warning is substantiated by concrete evidence. An evaluation conducted by the U.S. National Institute of Standards and Technology (NIST) explicitly found evidence of political censorship embedded within downloaded DeepSeek models, even when those models were operated solely on NIST’s own secure infrastructure. This finding directly undermines the assumption that self-hosting eliminates Beijing’s influence: "hosting those models in the UK wouldn’t, by itself, remove censorship," Burley emphasized. The implications for businesses and public services relying on such systems are severe, as it means embedding Beijing’s political restrictions into tools used for critical information processing and decision-making. Furthermore, the official policy governing DeepSeek’s online chatbot reveals it collects user prompts and uploaded material, processes and stores this personal data within China, and explicitly permits disclosure to Chinese authorities. Michael Wooldridge, a professor of AI at the University of Oxford, concurs that it is "not unreasonable to assume" that data inputted into such a chatbot could ultimately be shared with the Chinese state, posing significant risks to user privacy, commercial confidentiality, and national security.
The consequences for Western policymakers, particularly in the UK, are profound and urgent. The nation must resist the seductive pull of short-term savings offered by Chinese AI and instead focus on strengthening the resilience and security of homegrown and allied (notably U.S. and EU) artificial intelligence ecosystems. Continuing to rely on Chinese technology – whether in the form of electric vehicles or foundational AI models – risks accelerating the hollowing out of vital domestic industries, eroding technological sovereignty, and creating critical dependencies that Beijing can exploit. As the article concludes, the path forward requires rejecting the lure of cheap Chinese tech in favour of strategies that better "tame" U.S. AI while actively supporting secure, transparent, and democratically accountable AI development within the UK and its closest allies. The true cost of "free" AI from Beijing may prove far higher than any invoice.
https://www.theguardian.com/business/2026/oct/17/why-britain-should-reject-the-lure-of-cheap-and-scary-chinese-ai