In mid-January 2025, the UK Prime Minister, Keir Starmer, announced a step up in the UK’s AI policy by embracing nearly all the recommendations in a report by the tech entrepreneur Matt Clifford.
Clifford articulated four core principles to guide UK AI policy:
Government in policy, regulation and operations should be ‘on the side of innovators’.
Government should invest in becoming a great customer by harnessing its purchasing power: he cautions that this well is not easy, but requires real leadership and radical change, especially in procurement.
Government should actively ‘crowd in’ capital and talent: as the UK is a mid-sized economy, the government needs to proactively assist private sector build AI scale.
Government should stretch its efforts to build on UK strengths and catalytic emerging areas, such as in robotics and AI science.
Building AI infrastructure
Clifford urged for a substantial, economy wide investment in UK computing power, which would ‘reap benefits through increased economic growth, the reinvigoration of former industrial sites and ownership of critical strategic assets’, but also because this learning by doing would send a strong signal to global AI talent and investors that the UK was serious about AI.
The UK Government adopted the following specific recommendations to achieve this computing goal:
By 2030, 20x increase in computing power of the government-funded cluster of interconnected university and research computer, AI Research Resource (AIRR). Because ‘spreading large amounts of compute thinly will have little impact’, autonomous managers of the enlarged AIRR capacity will be appointed to allocate compute capacity in a strategic, mission driven manner.
Because even this enlarged sovereign compute capacity will still not be enough, the government will establish AI Growth Zones (AIGZ) in which privately funded data centres will be matched with energy generation facilities. The first AIGZ will be established in 2025, next door to the UK Atomic Energy Authority’s headquarters.
The UK intelligence agencies will work with the private sector to secure private sector AI compute facilities, including potentially to partition highly secure single platforms into government and private sector sectors.
The UK will seek to pool computing resources with like-minded countries to build global scale and reach.
Clifford recognised that access to high quality data was as important as large-scale computing power in building AI. The government accepted his recommendation to build a National Data Library by:
Rapidly identifying and releasing the top five high-impact government datasets to demonstrate the power of publicly owned data for AI training.
Strategically shaping what data is collected by government, rather than just making data available that already exists, with the actively seeking proposals from researchers and industry on data which could be collected for AI by government fiat.
Coupling compute allocation with access to proprietary data sets in packages which would attract researchers and developers to undertake AI development in the UK.
Actively incentivise and reward researchers and industry to curate and unlock private datasets.
Globally commercialising a copyright-cleared data set for AI training, drawing on content from content-rich UK institutions such as National Archives, Natural History Museum, British Library and the BBC.
Clifford recognises “the uncertainty around intellectual property is hindering innovation and undermining our broader ambitions for AI, as well as the growth of our creative industries”. The UK Government’s proposal for resolving this uncertainty is to implement an EU-style Text Data Mining right to give AI developers significant scope to train AI without compensating rightsholders.
Government as an AI customer
While there were some examples of innovative use of AI within the UK bureaucracy, Clifford characterised these as being small scale and in silos and that a more fundamental change was needed to ensure that government moved fast and learn things.
The UK Government has embraced Clifford’s recommended Scan > Pilot > Scale approach:
Scan more methodically and widely for opportunities and technologies by appointing AI leads for each government mission, creating a cross-government technical horizon scanning and market intelligence capability and two-way partnerships with AI vendors and startups, which would involve more actively pursuing joint use cases with the private sector for new AI tool development.
Pilot new AI more agilely by developing a faster, multi-stage gated and scaling AI procurement process to be applied consistently across government which would give individual agencies easier access to small-scale funding for pilots and only layers bureaucratic controls as the investment-size gets larger.
Scale successful pilots quicker and to greater scale by aggregating demand and use (e.g. across layers of government or regional authorities), establishing a centralised unit to provide technical and financial support to scale and mandating use of interoperable or open-source technologies.
Address private sector barriers to AI development
The UK Government accepted Clifford’s recommendations that it take a more enabling role in promoting private sector AI investment by:
Appointing AI Sector Champions in key industries like the life sciences, financial services and the creative industries to work with industry and government and develop AI adoption plans.
Replicate government-wide Amazon’s internal approach of open access by requiring all agencies’ AI-related documentation, data and functionality to be exposed through Application Programme Interfaces to which private sector companies could connect using their own tools.
Promote accelerated national adoption of AI by reach out beyond the AI-specific community, such as trade associations and SME organisations.
Clifford recommended establishing a new unit, UK Sovereign AI, with the power to partner with the private sector in AI development. UK Sovereign AI would have the mandate to bulldoze through barriers within government faced by private sector AI companies, such as consents and approvals, find suitable locations for data centres on government land, making government data sets available and get the intelligence agencies onboard.
Clifford also envisaged that UK Sovereign AI could make “direct investments into AI companies, including promising start-ups as well as joint ventures with other commercial partners [and] in exchange, UK Sovereign AI should ensure economic upside from, and influence on, governance of frontier AI for the UK”.
While adopting the recommendation for a new co-ordinating unit with the private sector, the UK Government was silent on whether it would have the mandate and funding for direct investment.
Conclusion
The UK Government’s announcement is part of a growing global policy shift from a singular focus on AI safety to promoting AI innovation, partly driven by a fear of falling far behind the US and China. While the Trump administration’s approach has been to revoke the Biden cornerstone AI Executive Order which set out AI safety requirements (an approach of government get out of the way of private enterprise), other governments are trying to find a new balance between AI safety and AI innovation goals. An emerging common thread in these non-US policy responses is a more activist, enabling approach by governments.
Read more: Prime Minister sets out blueprint to turbocharge AI
Peter Waters
Consultant
