UK’s AI regulation

In the UK, the government has adopted an agile, pro-innovation, sector-based approach to AI regulation. Avoiding disruption to responsible AI innovation is the government’s rationale for not rushing to legislate. A non-statutory approach could offer better adaptability.

In March 2023, the government published the AI White Paper. This set out a cross-sectoral, principles-based and regulator-led approach, with light-touch government involvement. This means the government has looked to empower existing UK regulators (including the Information Commissioner’s Office (ICO), Competition and Markets Authority (CMA), Financial Conduct Authority (FCA), Medicines and Healthcare products Regulatory Agency (MHRA) and Office of Communications (Ofcom)) to develop flexible and proportionate guidance to address the use of AI in their sectors, with a focus on the principles of:

• safety, security, robustness;
• transparency and explainability;
• fairness;
• accountability and governance; and
• contestability and redress.

To this end, the regulators were asked to report on how they are developing the proposals of the White Paper and their strategic approach to AI, the responses to which can now be found here. These updates could inform the government’s adaptive approach to AI, and may lead to a statutory duty being placed on regulators to have regard to the above principles.

Work has been underway to create a central government AI function to coordinate policy across sectors and promote information sharing between regulators. Investment (in the millions) has also been made to upskill regulators and support innovation in the AI market (though Labour has since scaled back certain funding that had been promised for AI projects).

The government recognises the importance of gap analysis to identify areas which require coverage or possible future regulation. This may be needed in relation to “highly capable general-purpose AI systems” to manage safety risks if this is deployed across multiple sectors (these are foundation models trained on huge datasets which can perform a wide variety of tasks, some to “superhuman” levels).

The UK has also sought to position itself as an international leader in providing a safe and trustworthy framework for AI development, having:

• hosted the first AI Safety Summit in November 2023. The EU and 28 countries (including China and the US) signed the Bletchley Declaration that seeks to develop AI in a safe, responsible and human-centric manner;
• established the first AI Safety Institute to provide advanced AI safety for the public good;
• co-hosted the AI Seoul Summit in May 2024. Further agreements to AI safety were announced, including the creation of a global network of safety institutes and commitments to develop shared risk thresholds for frontier AI; and
• signed the ‘AI Convention’ on 5 September 2024. This is the first legally-binding treaty to implement safeguards to AI, including provisions for the protection of human rights, democracy and the rule of law.

It’s possible that the UK might still legislate for AI, particularly given the change in government. The UK has an Artificial Intelligence (Regulation) Bill (Private Member’s Bill brought in House of Lords), which has since been passed and sent to the House of Commons. In July 2024, the King’s speech also announced the Government’s plan to introduce AI-specific legislation. This built on Labour’s manifesto pledge to ensure the safe development and use of AI models, including the introduction of binding regulation on certain companies that develop the most powerful of these models.