Should the UK build its own Large Language Models?

Large Language Models (LLMs) like OpenAI’s GPT, Google’s Gemini, and Anthropic’s Claude are transforming the way businesses operate, how public services are delivered and how knowledge is accessed. They’re already embedded in business operations across a range of sectors, from healthcare diagnostics to customer service. But they’re also raising questions about sovereignty, safety and strategy and have prompted a key question for UK policymakers, researchers and businesses alike – should the UK invest in building its own LLMs?

The case for a UK-built LLM

There are several strategic reasons for the UK to pursue its own LLM capabilities. 

First is digital sovereignty. Most of today’s leading AI tools are developed and operated by a handful of US-based tech giants. Relying on foreign systems within critical infrastructure, healthcare systems and financial services means that the UK has limited control over how these tools evolve, how they’re trained, how the data is handled or whether they align with our laws, values and strategic goals. This has the potential to create legal, ethical, economic and geopolitical vulnerabilities.

Economically, investing in LLMs could support innovation across a variety of sectors, from financial services and education to defence and life sciences. Domestically developed LLMs could strengthen productivity, unlock new business models and generate billions in value.

UK-built LLMs are also a talent strategy. The UK has world-class AI researchers and engineers, many of whom have played key roles in developing the models that are now deployed overseas, and many of whom have been attracted abroad by better-funded labs and global tech firms. A well-supported commitment to LLM development could help to retain and attract that talent, avoiding a brain drain, creating thousands of high-skilled jobs and supporting the growth of UK-based AI startups and research hubs to ensure we remain competitive.

Finally, we should consider geopolitics. The UK has positioned itself as a global voice on AI governance and ethics, hosting the landmark Bletchley Park AI Safety Summit in 2023, and aspires to shape international regulation. Leading by example with its own model would reinforce the UK’s credibility on the world stage and enhance its influence. Without it, we risk being sidelined in future policy debates.

A risky business?

However, the scale and cost of developing an LLM requires substantial investment. Training a cutting-edge model can cost hundreds of millions of pounds (often more than the entire national R&D budget in other sectors, and that’s before deployment, safety testing and updates are factored in) and there’s no guarantee of commercial or technical success. The fact is that OpenAI, Meta and Google have many years’ head start, seemingly unlimited finances and vast data resources, so the UK is already starting from behind and, for a relatively small market, the commercial return is uncertain.

There’s also the risk of duplication. Building another ‘GPT-style’ model may not add value if it simply duplicates what’s already there and, in fact, might waste resources and could distract from areas where the UK already leads. Many experts argue that the UK should focus on niche or domain-specific tools in which it can differentiate and lead.

LLMs also carry significant ethical concerns. They can amplify bias, spread disinformation and have the potential to be misused, and that’s before we scrutinise the environmental issues. Any UK investment would need to be built with a strong regulatory and governance framework which contains robust safeguards, built-in ethics, transparency and accountability to guarantee public trust.

Taking a strategic approach

Rather than trying to compete with Silicon Valley and build a British version of ChatGPT, the UK could take a more pragmatic approach, which might involve:

• Investing in open-source and public-benefit LLMs, such as the UK’s AI Research Resource, and developing models tailored to the UK’s strengths – medical, legal or defence applications – that prioritise transparency and accountability  
• Funding of national AI infrastructure, like compute resources and high-quality datasets which support long-term innovation in universities and start-ups
• Strengthening international partnerships and collaborating with European and Commonwealth allies to pool resources, share standards, align regulation and increase impact
• Prioritising AI skills and talent attraction and development to create an environment where the UK is the destination of choice for world-class talent who can focus on responsible, human-centred AI deployment.

In conclusion

Here at CBSbutler, we don’t think that the UK should aim to build a general-purpose LLM simply to keep up with global tech giants. However, we believe it should invest in AI capabilities, selectively and strategically, that reinforce our innovation economy, talent pipeline and global standing, and focus on areas where we can lead, to ensure our values shape the future of technology. In this way, the UK can play a meaningful role in influencing the next phase of AI, without having to replicate what others have already done.

If you’d like to discuss any of the issues here or want to contribute towards the debate, call us on +44 (0)1737 822 000 or fill in the form here.