The report finds that the UK’s adoption of AI is a key part of state strategy that appears to be principally motivated by improving economic performance and bolstering innovation.

The paper highlights risks resulting from how AI systems are being developed, procured and deployed across the public sector. These include:

• A lack of scrutiny of AI systems being procured for the benefit of government departments and for the powerful companies that provide them, but not necessarily for the benefit of the general public.

• Government partnerships with private companies risk abuses of power and state capture through Big Tech successfully lobbying for further use of their technology across state organs.

• AI is being procured in a way that stifles competition and innovation, and risks government departments being locked into agreements with mostly large US technology companies. This position risks the UK’s sovereignty in times of geopolitical uncertainty and economic coercion.

• AI systems can produce and perform policy unseen, leaving the public in the dark about how life-changing decisions are being made.

• There is minimal indication that AI development is being undertaken with reference to laws on data protection, equality, and human rights. Even if a court rules against the use of a specific AI system, legal judgments cannot address systemic issues with the technology itself and how its use contributes to further societal harms.

Executive Director of Open Rights Group, Jim Killock said:

“The Government is putting the interests of Big Tech ahead of the British public in spending billions on AI systems, which it claims will create efficiencies.

Not only are efficiency claims likely to prove exaggerated, the Government risks further locking the public sector into using a handful of dominant US technology companies, stifling competition, undermining innovation and creating long-term risks for the UK’s security and resilience. At the same time, there are no effective and coherent mandatory rules to ensure these systems respect human rights.

The public is being kept in the dark about how systems affect them. We urgently need AI regulation that puts transparency, accountability and the public interest first, protecting human rights and ensuring AI works for all of us.”

Dr Richard Mackenzie-Gray Scott, the paper’s author, said:

“With the recent change in government, there is a new opportunity to align digital technology policy with human rights.

We need to start being honest about what makes new and emerging technologies possible. What the costs of their development, maintenance and use amount to in different contexts, and whether and how these costs are offset by any related benefits. There is considerable false and misleading information being spread about “AI”.

It is also hard to keep track of the promises and predictions, especially from cash furnace companies. But those responsible for governing need to focus on the harms that exist now, on the risks that exist now, and on ensuring that innovation for the sake of lining pockets and increasing power is replaced by innovation that contributes to social value.

The core of such a change should be human rights, lest the illusion of progress persist.”

The UK’s public spending on AI rocketed by over 800% from 2018 to 2025, with a current total of approximately 1,500 contracts worth a reported £3.35 billion.

The UK has already entered into a number of partnerships with companies including Amazon, Google, Microsoft, OpenAI, Oracle and Palantir, with the purpose of attracting and securing foreign investment in exchange for public sector authorities using technologies provided by these companies.

The government claims that AI in the public sector could help save anywhere in the range from £45 to £87 billion in annual expenditure. But claims that AI will deliver dramatic efficiency and gains currently remain unproven, while costly errors, legal challenges, and the need for ongoing human oversight could leave taxpayers worse off.

Investing in lower-cost, open-source alternatives could deliver better value without exposing the public to the risks of deploying unreliable AI systems at scale.