Does Britain need a more autonomous approach to AI development?

In the 19th century, innovation and industry were central to Britain's economic power and imperial expansion. Machinery, capital and industrial expertise flowed across an increasingly interconnected world, contributing to profound changes in global economic development.

Two centuries later, this has ceased to be the case. Britain has since pivoted to a predominantly service-based economy, with service industries accounting for 81% of total economic output. Having shifted away from large-scale industrial production, its comparative advantage rests on cultivating intellectual capital, specifically research, finance and professional expertise.

This strategy worked well so long as the UK could rely on a close economic and technological relationship with the US. It meant Britain could cede technological development to their new national pastime of outsourcing, whilst trusting the US to create and share future innovations. Likewise, it proved a happy arrangement for American big tech too, as Britain offered a safe and reliable market for investment, whilst London became the pre-eminent hub for European operations.

However, this arrangement ultimately relied on American technological dominance being embedded within a broadly multilateral order, which AI has since complicated. OpenAI and Anthropic’s frontier models are becoming strategic assets with implications for national security, economic competitiveness and state capacity. The US government views technological leadership as national power, especially as Donald Trump has framed AI development within an increasingly confrontational competition with China.

Britain cannot realistically compete at this level, so adopting American AI is pragmatic and in line with its established technology importation strategy. The issue, however, is that Britain has built little technological leverage should American interests diverge from its own – a once unforeseeable occurrence. Yet, the problem is no longer entirely hypothetical.

Earlier in September, Anthropic reportedly refused to submit its latest model, Mythos 5.1, to the UK’s AI Security Institute for testing, an unprecedented departure in their usually cordial relations. If AI companies, or the US government, were to restrict access to groundbreaking frontier models to the UK, it would leave Britain technologically weakened and strategically disadvantaged. The same applies to data. If critical information and public services in areas such as healthcare, education and security become dependent on foreign AI systems, London risks compromising another layer of its autonomy and security.  

The UK government has recognised the need to develop AI sovereignty, but its efforts lack direction, scope and strategy. The Department for Science, Innovation and Technology (DSIT), created under the Sunak Ministry to streamline AI development, has been disbanded after just three years. Its responsibilities have instead been distributed across other departments, raising questions about whether the specialist institutional capacity required for AI policy has been preserved.

Forming the Sovereign AI Unit (SAU) showcased ambition, but with investment in eleven domestic start-ups to the tune of £500 million, it may fail to compete with even France, which has committed over £100 billion to its own programmes. Similarly, the Isambard-AI at the University of Bristol, funded by DSIT to fuel the SAU, is Britain's most powerful supercomputer, but only 13^th in the world, behind states such as Finland and Italy.

Aside from the government’s strategy, Britain also currently lacks the capacity to systematically convert its considerable intellectual capital into technological ownership. This is most notable with Google’s acquisition of DeepMind. It proved successful; immense capital allowed DeepMind to operate at a scale unprecedented for British-backed companies.

However, it became symptomatic of Britain’s position: its ability to generate valuable intellectual capital without always owning the capital or institutional capacity required to retain and scale it domestically. Since then, OpenAI has established its largest non-US research hub in King’s Cross; Anthropic collaborated with the DSIT to develop and pilot an AI assistant on GOV.UK and Google has ventured into data centre construction whilst partnering with the government on AI training for civil servants.

So, what is Britain's solution? Well, it cannot pursue technological autarky. Not only does it lack the capital, energy, manufacturing capacity and domestic market to reproduce the vertically integrated technological ecosystems constructed by China and the US, but Britain’s economy has long benefitted from an international position within trade, finance, research and investment networks.

Rather, it needs to distinguish between what is commercially beneficial and what is strategically important. It is fine to import consumer electronics, be embedded within a Western semiconductor supply chain, and integrate foreign software in business and some aspects of government. It is dangerous to have no meaningful domestic capability in technologies which the wider economy and state increasingly depend upon. The issue, then, is not dependence itself, but the kind of dependence Britain is willing to accept.

Britain has the potential. The London-Oxford-Cambridge corridor provides intellectual capital; the fruits of this capital can be seen in DeepMind, ARM, Wayve and Isomorphic Labs. What it needs is a stronger mechanism to connect ideas with finance and commercialisation. Government can provide the infrastructure and seed funding where private investment is unwilling to bear the initial risk. Pension funds, institutional investors and the City can provide the scale to grow this promising research into globally competitive companies.

The result is not simply a case for spending more public money but also mobilising Britain’s considerable financial capacity behind its technological capacity. If Britain can connect its financial strength with the research concentrated along the London-Oxford-Cambridge corridor, it may possess something closer to a complete ecosystem for technological development.

 This, rather than technological autarky, is what strategic sovereignty should mean in the context of Britain’s AI development. Britain doesn’t need to manufacture every semiconductor, train every frontier model, nor construct an entirely independent technological ecosystem. It just needs to identify the technologies upon which its future prosperity and security depend, retain meaningful capabilities within those areas, and ensure that the country has alternatives when foreign interests diverge from its own.

The objective, therefore, is not to build everything ourselves. It is to ensure that Britain is never so dependent upon someone else’s technology that it loses the capacity to make its own decisions.

Image: Wikimedia Commons/Redbannana

Licence. 

No image changes made.