Keir Starmer and AI: potential positives but possible pitfalls

In a week in which Donald Trump’s impending inauguration helped drive up the cost of government borrowing in the UK, an optimistic speech on AI from Keir Starmer provided a convenient replacement for the one Rachel Reeves had been due to deliver on growth. 

In many ways, the prime minister’s launch of Matt Clifford’s AI Opportunities Action Plan represented continuity from the AI-enthusiasm of Rishi Sunak’s administration, but Starmer sought to differentiate Labour’s approach to AI from what – he argued – was the Conservative Party’s more precautionary, safety-focused stance. It was a positive vision, and one which could bring benefits for the civil service – though a sceptical public will need convincing, especially if, or rather when, pilot schemes fail.

Trump’s presidency will shape the UK’s approach to AI regulation

The PM’s speech – in which he committed to ‘take forward’ all 50 of Clifford’s recommendations – focused on the positives. Starmer set out his government’s ambition to support the development of a distinct AI ecosystem in the UK by increasing the country’s capacity to develop and benefit from these new technologies by facilitating access to talent, energy, data and computing power. 

In terms of regulation, the PM argued that the UK will be able to carve out a distinct position on AI regulation between the regulatory behemoths of the US and Europe. This is an ambitious position. Only as the reality of Trump’s second term emerges will it become clear whether it can be sustained or whether the UK will in fact become a pawn in a high-stakes game of regulatory chess between the major tech powers. 

While the PM sought to move the AI narrative away from safety and assurance and towards opportunity, it was notable that the journalists who attended the speech approached the subject with a very different mindset. Almost all questions to the PM focused on the risks and disbenefits of the new technology. 

The caution of the media reflects the concerns of many members of the public about AI, and the challenge for government of pursuing its benefits while managing the risk of undermining trust in the state when things go wrong. Because Starmer’s vision was not just about establishing conducive conditions for private sector AI companies to flourish – creating jobs and prosperity. It was also about the benefits that he anticipates AI can deliver for the public sector, particularly in terms of productivity but also problem solving. The Spending Review will be the first opportunity to see whether Starmer’s AI enthusiasm is backed up by serious financial clout – though a chancellor reportedly looking for further savings may require persuading, or telling, by the PM.

Ministers need to brace themselves for AI pilots that fail

The government’s AI Action Plan includes a number of recommendations designed to tackle perennial government problems that will have to be resolved if the potential benefits of AI for government are to be realised. These include a commitment to the “creation of a technical senior civil servant stream, benchmarking of internal AI-related role pay to at least 75% of private-sector rate and a technical AI recruitment screening process." The Institute’s previous work has highlighted the shortcomings of civil service recruitment, pay and talent development in relation to specialisms such as data and AI, so these are welcome commitments. 

More broadly across the public sector, the adoption of a ‘scan > pilot > scale’ framework is designed to address the problems government has experienced with scaling up pilots in specific parts of government to leverage wider benefits. Ideas such as the Government People Service’s tool to optimise job descriptions, or the ONS’s tool to improve the analysis of free text survey responses, have the potential to generate much wider benefits for government if rolled out more widely. Ministers will need to make themselves comfortable with the reality that some pilots will fail – and that they will need to communicate these failures to the public as a feature, rather than a bug, of the approach they have chosen to adopt. 

Clifford will now advise government on the implementation of his report – and it is welcome that he will remain involved. In the coming months it will become clearer whether this burst of positivity from the PM translates into a real commitment to make AI central to his government's missions and milestones, and to rewiring the British state.