Britain is being urged to develop its own sovereign artificial intelligence (AI) capabilities to reduce growing dependence on the United States, experts say, amid concerns about the risks of relying heavily on foreign technology providers. Recent developments have highlighted vulnerabilities associated with using AI systems controlled by companies based outside the UK, particularly those in the United States.

Over the past several decades, the UK has relied extensively on American-made technologies—ranging from Trident nuclear missiles and F35 fighter jets to Nvidia chips, cloud services, and GPS satellites—as part of the longstanding “special relationship” between the two countries. However, recent events have underscored the potential downside of this dependence, with Washington’s ability to restrict access to critical technology emerging as a significant concern. A notable example cited is the recent throttling of Starlink satellite internet service to Ukraine, which demonstrated how access can be curtailed.

In the AI sector, Britain’s public and private sectors currently use tools developed by major U.S.-based firms such as OpenAI and Anthropic. These companies operate “opaque” AI models whose internal workings are not publicly available, meaning users pay for service access but do not own or control the underlying technology. This contrasts with “open-weight” AI models, which are transparent, downloadable, and can be hosted on private servers, allowing users to modify and customize the technology.

Jensen Huang, CEO of Nvidia, the world’s leading chip manufacturer, recently told British officials that reliance on closed systems operated by Silicon Valley companies is “not sustainable.” He advocated for the adoption of open AI models that offer greater independence and flexibility. Huang’s comments came during a visit to London where he met with government leaders, including the prime minister, to discuss technology strategy.

The importance of a sovereign AI system was illustrated earlier this year when European users lost access to Anthropic’s Fable model due to a U.S. Commerce Department directive restricting non-American access. Although the immediate impact was limited by AI’s still-nascent role in public administration, the incident serves as a warning about potential future disruptions if Britain remains dependent on foreign AI providers.

The UK government has introduced a suite of AI tools collectively named “Humphrey,” which has already shown promise in public sector applications. For example, it was used to analyze 50,000 submissions during a water industry consultation in a matter of hours, a task that would have otherwise taken civil servants days. Another AI application sped up the digitization of records for local planning authorities. Despite these early successes, Humphrey currently relies on a mixture of AI models from American firms, which leaves the system vulnerable to external restrictions or cuts in service.

Advocates for a sovereign AI argue that Britain can leverage advanced open-weight models produced by Chinese companies like DeepSeek and Alibaba, as well as European and American contributors such as Meta, Google, and the French AI lab Mistral. By hosting these models domestically and adapting them as needed, British engineers could produce AI tools under national control. This strategy would enable the country to maintain data sovereignty and avoid dependence on foreign governance or censorship rules, such as those imposed on Chinese models.

Given that Britain has limited indigenous AI infrastructure since the acquisition of DeepMind by Google, open-weight models offer a practical path to building a domestic AI capability without starting from scratch. Establishing a dedicated team—potentially spun off from intelligence agencies like GCHQ—could oversee the evaluation, adaptation, and enhancement of available open models to create British versions suited to government and commercial use.

Ultimately, while complete independence from American technology remains challenging, developing a sovereign AI system could reduce reliance on foreign providers and ensure greater resilience for Britain’s critical digital infrastructure as AI becomes increasingly integral to public sector operations.