Britain’s AI Soft Power: What the AISI reveals about the limits of British influence

Ruby Lee, Intern 
21/09/2026

The UK AI Security Institute (AISI) illustrates a wider problem facing Britain, its influence over frontier AI depends entirely on the goodwill of companies that it has no formal leverage over. Established in November 2023 by former Prime Minister Rishi Sunak, the AISI has become a world-leading institution for testing the capabilities and security risks of advanced AI. Rather than competing with California’s technological dominance, Britain projects influence through research leadership and international cooperation, a strategy that grants British investors and developers pre-release access to frontier models. However, this depends on developers choosing to submit their models for voluntary testing.

The AISI has continuously provided Britain pre-release access to advanced AI models for safety and security evaluations, maintaining British early access to the largely US-developed advanced models. This influence is dependent on voluntary cooperation, exposing Britain’s precarious access to these models. This reliance was exposed when it was reported on 9th September that Anthropic declined to submit their latest model, Mythos 5.1, to the AISI for pre-release testing. By contrast, the AISI has been granted pre-release access to every Anthropic frontier model since 2023. This decision follows a broader climate of AI protectionism among American developers, following the Trump administration’s temporary restriction of access to Anthropic’s Fable and Mythos models for foreign nationals. While not formally linked, these events demonstrate that Britain’s access to frontier AI models, remains firmly outside of British control, in the hands of private companies and the US government. Soft power can create access and influence, but it cannot compel either, and Anthropic’s exclusion of the AISI is a reminder of Britain’s insecure access to this technology.

Britain’s AI strategy occupies a middle ground between the EU’s regulatory framework and Trump’s complete rejection of AI controls. The AISI contributes to this position, by keeping Britain relevant to international AI governance, without competing with American frontier companies. This demonstrates how the AISI keeps Britain influential in an industry where the underlying technological innovation is concentrated abroad.

The limits of Britain’s position carry direct economic consequences, and raise the question, can Britain realistically steer a middle course between the US and the EU? Potential future EU single market access is likely to come bound by EU AI regulation, while a US trade deal could involve more access for American technology companies. Does Britain have the economic weight to continue to occupy this position, and at some point will Britain have to favour one side? Reliable access to frontier models is significant for economic growth, as technology is one of the few sectors of the British economy expanding year-on-year, and can compete internationally. That the government has commissioned a report quantifying the economic and competitive costs of losing frontier AI access suggests a state managing dependency that it cannot control, rather than one setting the terms. Losing access would also threaten Britain’s standing as a research hub, since with the elite Golden Triangle universities rely heavily on frontier AI models for research and spin-out activity. Meanwhile, Bank of England Governor Andrew Bailey warned G20 finance ministers in August about the cyber security implications of frontier AI for global financial stability. Therefore, Britain faces a challenge, to manage the sufficient scrutiny of frontier systems to protect its reputation as a world-leading regulator, while preserving the access on which economic growth is dependent on.

Indeed, the AISI delivers influence Britain could not obtain through technological dominance alone, but whether soft power is sufficient to keep Britian relevant while the underlying technology is controlled by others is the crux of the matter. Anthropic’s decision exposes the vulnerability of a regulation model based on cooperation, while the US export controls demonstrate the additional leverage held over British interests. For British policymakers, this turns AI soft power into a question of international reputation and economic sovereignty, how can Britain continue to shape an industry when it cannot guarantee access to the technology?