Two weeks ago, Kanishka Narayan, the Minister for AI and Online Safety, declared that “Britain will now become the home of global open-source AI talent”. This is the right level of ambition; the UK can be an open-source AI superpower capable of driving and influencing the global ecosystem. But this ambition must be matched with action.
To support this goal, many have called for Britain to develop its own “sovereign” ChatGPT-equivalent model. This is a mistake. Britain cannot compete with the US or China to lead at the AI frontier because of the sheer scale and capital required. More importantly, it does not have to. Instead, the country should support and build a new centralised capability – the UK Open-Source AI Lab.
Britain already has a track record of success in building and investing in many of the foundations necessary for sustaining a vibrant and innovative AI ecosystem. The new “AI Growth Zones” help ensure access to computing power, and the AI Research Resource helps drive innovation by providing compute access to the strongest British startups and academics.
Now, the country must focus on continuing to develop a world-leading national ecosystem that can dynamically capture the economic value of AI and build the tools, industries, and public services of the future. As the Tony Blair Institute for Global Change’s recent paper Open Source: How Middle Powers Can Build Influence in the Age of AI lays out, the UK can build sovereign talent capability, drive AI adoption across public services and the economy, improve security by maintaining the critical infrastructure we all depend on, and project AI leadership internationally – all by leveraging open source.
The Case for Open Source
Open source is already central to the UK and the global economy. Software that is “open sourced” underpins 70 to 90 per cent of modern digital systems globally. If everyone suddenly stopped contributing to open-source development, long-run GDP in the average economy would be 2.2 per cent lower. In the UK specifically, open-source software contributed an estimated £46.5 billion in direct and indirect economic value in 2020 – that’s three times the average annual contribution of the UK aerospace industry.
Open source was already critical to the digital era. It will only become more important in the age of AI, for two reasons.
First, like open-source software, open-source AI models can be adapted to build specific tools by training on new, specialised data. This is a lot cheaper than building from scratch and allows smaller players to innovate alongside wealthy AI labs in the US and China.
Second, without open-source software, we would not have frontier AI as we know it – even the most capable models today heavily rely on open-source foundations.
For both reasons, there is a clear opportunity for the UK: because open source is foundational to AI, building capability within these foundations will deliver a greater return on investment than trying to own the frontier.
This is a durable strategy and a safe bet. While open-source AI models are unlikely to ever surpass the capabilities of closed-source frontier competitors, they are rapidly catching up. And yet even if, in future, proprietary models far outstrip open-source alternatives, once again, there are important reasons why open source would remain critical:
First, innovation requires adaptation – and open source enables that. Especially in high-stakes sectors (think AI for health-care diagnostics), people ultimately need tools built specifically for their applications. By enabling adaptation, open source can allow the UK to capture some of this “downstream” innovation.
Second, smaller systems will likely beat big ones for certain tasks due to the demand for efficiency. A scientist using machine-learning techniques to organise a huge astrophysics data set does not necessarily need a “superintelligent” system that costs hundreds of thousands of pounds to run. They need an efficient model trained on their specific data set, which is why the ability to “distil” open-source models into smaller, specialised ones is so important.
Finally, interoperability will be as important as scale. As “agentic” systems – with different AI agents being linked together – emerge, modularity will become increasingly important. The ability to combine and swap components improves capability and reduces vendor lock-in. Enabling that modularity requires open standards, reusable components and shared benchmarks.
For the UK, openness is therefore a key policy lever: a way to build capability and retain agency without sinking public money into low-probability gambles.
The question is how the UK can now organise itself to capitalise on the open-source prize.
What the UK Open-Source AI Lab Would Do
An Open-Source AI Lab would create a centre of gravity for the UK’s AI ecosystem and make the country the world leader in open source – much as the AI Security Institute (AISI) has allowed the country to lead in AI safety. Specifically, the lab would:
Develop open tools for government and public services. Build reusable adapters, safety layers, evaluation harnesses and deployment pipelines that departments can share, reducing duplication and strengthening internal capability.
Distil and fine-tune open-foundation models for priority UK sectors. Adapt these models for use in sectors like health and AI for science – leveraging key UK data sets.
Unlock and curate high-value data. Work with existing bodies like the Sovereign AI Unit to build strategic data sets in high-priority sectors and research areas, and ensure they are usable by UK startups and researchers.
Maintain critical open-source infrastructure. Treat open-source tooling and maintenance as critical infrastructure, supporting maintenance to build the UK’s digital resilience.
Lead international collaboration. Coordinate with allied initiatives – such as Germany’s Sovereign Tech Agency and the proposed EU Sovereign Tech Fund – and work with partners like Mozilla and GitHub to shape shared standards and benchmarks.
The lab would not only build tools and maintain critical software infrastructure. It would be a domestic touch-point for a dynamic open-source ecosystem, and would – along with the AISI – give the UK a stake in leading the global AI debate. International coordination will only become more critical as the US and China race ahead, and as middle powers look for trailblazers to lead the rest of the world in global AI cooperation.
Crucially, the UK does not need to start from scratch. The government’s existing in-house Incubator for Artificial Intelligence (i.AI) has already shown that Whitehall can build practical, reusable AI tools using open, product-led engineering. The logical next step is to build on that foundation by evolving i.AI into a dedicated Open-Source AI Lab, with a clearer cross-government mandate and long-term funding.
Structurally, the lab would require a small technical core but also flexible fellowships to attract high-impact open source contributors and maintainers – working alongside existing UK efforts to attract open source talent. It would ideally report to a high level – either the Minister for AI and Online Safety or the Chief AI Officer in Number 10 – reflecting its cross-cutting, national importance.
Externally, the lab would also need deep links to the open-source community, including companies and organisations such as Mozilla, Hugging Face, GitHub, the Linux Foundation, OpenUK, and the AI Security Institute, as well as individuals on the ground who have an outsized impact on the open ecosystem. This would ensure that public investment compounds rather than duplicates existing efforts, and would also lay the basis for more formal public–private partnerships.
Building a World-Leading AI Ecosystem
The UK has a once-in-a-generation opportunity to position itself as the global leader in open source and shape the open foundations of the fast-approaching AI economy.
Establishing the UK Open-Source AI Lab is the first step. At its core, this approach is about rejecting passivity, building the UK’s domestic capability to be a better innovator and adopter of AI, and having a stake in shaping the technology’s future both at home and abroad.
The path to gaining sovereignty, building leadership and actively influencing the impact of AI, rather than passively experiencing it, lies not in a model but in an ecosystem. Fortune will favour the “middle power” that recognises this first and takes the necessary bet. It should be the UK.