FTI Consulting UK Public Affairs Snapshot: AI Opportunities Action Plan: Starmer’s high-stakes AI gamble

The AI Opportunities Action Plan was published last week by Matt Clifford, a tech entrepreneur and the Chair of the Advanced Research and Invention Agency (ARIA). Commissioned by the Secretary of State for Science, Innovation and Technology last July, it makes a bold case for Britain to bet big on advanced homegrown AI to deliver growth. According to the IMF, the prize on the table is a productivity boost of 1.5% a year, which the government estimates will amount to an average of £47 billion to British economy each year over a decade.  

Reflecting Clifford’s start-up background, the plan calls for the government to embrace risk and take an entrepreneurial approach to statecraft. It argues that to become one of the biggest winners from AI and drive national renewal, Britain must go beyond simply building AI infrastructure and encouraging the adoption of AI – it must become a player, or an “AI maker.” If fully implemented, this would see a significantly increased role for the state in developing AI.  

The Prime Minister, Keir Starmer, echoed this message. He stated that the government will be “emboldened to take risks as our brilliant entrepreneurs do”, adding that the state “shouldn’t just focus on safety and leave the rest to the market”. Starmer was openly contrasting his approach with Rishi Sunak’s agenda, which was criticised by industry for focusing too heavily on AI safety over AI opportunity. 

Nonetheless, Clifford’s plan praises Britain’s pro-innovation regulatory framework – the progress of which was started under Sunak – as a key strategic advantage over more aggressively regulated jurisdictions. This provides reassurance to industry that the upcoming consultation on the AI bill will continue this approach. 

The plan makes 50 recommendations across three pillars, all of which were accepted  by the government, which pledged to put the “full weight of the British state” behind AI. The first focuses on laying the foundations to enable AI by growing Britain’s compute 20-fold by 2030. This will be achieved by encouraging domestic data centres to be built in “AI Growth Zones”, unlocking the power of public data sets, developing pro-innovation regulation and incubating AI talent in the hopes that “one individual founds the next DeepMind or OpenAI”.

The second pillar emphasises public sector transformation through AI adoption, including a recommendation for the government to act as a “customer” of AI to drive broader private-sector uptake.  

The third builds on pillars one and two, arguing that the government needs to foster homegrown challenger companies that can compete globally. To support this, the plan recommends creating a new government unit – “UK Sovereign AI” – with the power to partner with the private sector to maximise Britain’s stake in advanced AI.  

The plan’s ambition has been well received by industry, with 10 leading AI companies pinning their quotes to the government’s accompanying press release. Paris-based firm Minstrel AI’s decision earlier this month to name London as the home of its first European office outside France demonstrates that Britain is putting out the right signals to business. 

However, despite the generally warm reception, copyright campaigners remain concerned with the government’s proposal to reconcile copyright and AI development by extending the text and data mining regime. Striking the right balance between protecting Britain’s creative industries and fostering AI innovation will be delicate for the government to navigate.  

However, in the face of a sharp rise in Britain’s borrowing costs, will the Chancellor be willing to foot the upfront costs required? Matt Harris, senior vice-president of Hewlett Packard Enterprise, argues that even with help from the private sector, the government’s compute targets will require “billion-pound investments” from the Treasury. 

Last summer, the newly installed Labour government shelved plans to build an exascale supercomputer at Edinburgh University, arguing that the £800 million funding promised to it by the previous Conservative government “never existed”. And while there are reports that suggest a revival of the Edinburgh supercomputer in the upcoming compute strategy, nothing has been confirmed yet. 

Nevertheless, this incident, particularly stinging as Edinburgh University had already spent £30 million in preparations, casts doubt on the government’s willingness to commit the necessary funds for Clifford’s proposals to be realised in full. Unlike the low direct costs of introducing AI safety regulation, Clifford’s plan will have to compete with other core commitments that require significant government resources and funding.  

As we approach the Spending Review in June, whether or not the Chancellor is committed to financing the plan will become clear. Ultimately, the level of funding the government puts behind its ambitions will serve as a litmus test of its belief in AI as the answer to its economic problems.