The UK does not regulate artificial intelligence as a technology. It regulates the activity in which the technology is used, through the bodies that already oversee that activity. That choice was made deliberately in 2023 and has survived a change of government. What it has not yet produced is legislation.
The approach chosen in 2023
The government set out its position in a white paper, A pro-innovation approach to AI regulation, published in March 2023. It proposed that AI would continue to be overseen by existing regulators covering specific sectors, naming Ofcom for communications and Ofgem for energy. A context-based approach, applied by specialist regulators, was preferred to a single regulatory function with uniform rules.
The paper sets out five cross-sectoral principles that regulators are expected to apply: safety, security and robustness; appropriate transparency and explainability; fairness; accountability and governance; and contestability and redress. The government described these as values-focused, guiding how existing regulators respond to AI risks rather than creating a new authority to handle them.
The framing is the substance here. The white paper treats AI risk as arising in a context, and it concludes that the rules which matter are the ones already governing that context.
The paper states the choice in as many words. It says the government will not put the principles on a statutory footing initially, on the reasoning that rigid and onerous legislative requirements could hold back innovation and reduce the ability to respond quickly to future advances. The principles were to be issued on a non-statutory basis and implemented by existing regulators, using their domain expertise to tailor how they apply.
It also sets out the condition for changing that. The paper anticipates introducing a statutory duty requiring regulators to have due regard to the principles, when parliamentary time allows, and adds that it will not introduce such a duty if monitoring shows implementation is effective without legislation. The non-statutory approach was therefore designed to be tested against its own results.
One thing the framework explicitly does not create is a new regulator. The paper states that its central support functions will be provided from within government without establishing an AI regulator, and that those functions will neither replace nor duplicate regulators' work. The coordination layer it describes covers monitoring the framework's effectiveness, assessing economy-wide AI risks, horizon scanning, supporting sandboxes, and promoting interoperability with international frameworks.
The paper was also candid about why speed was being prioritised. It cites Sir Patrick Vallance's Regulation for Innovation review, which identified a short window for government intervention to create a clear regulatory environment, and it records that the UK is frequently ranked third worldwide across measures including AI investment, innovation and implementation. It notes that industry repeatedly emphasised that consumer trust is key to the success of innovation economies.
It also named the cost of the status quo. Because AI is a general-purpose technology whose uses cross regulatory remits, the paper states that it is currently regulated through a complex patchwork of legal requirements, and records industry feedback that the absence of cross-cutting regulation creates uncertainty and inconsistency that can undermine business and consumer confidence.
What that asks of existing regulators
The paper works through an example: an insurer using AI to set premiums. That activity would be subject to data protection, equality and general consumer protection law, and possibly sectoral rules such as the Financial Services and Markets Act 2000. Depending on how the system is used, it might also engage medical device regulation.
The same passage identifies the difficulty. The paper states that it can be difficult for a company in that position to identify which rules are relevant and to apply them with confidence, describing a lack of cross-cutting principles and limited system-wide coordination. That is a description of the compliance problem from the perspective of the regulated party, and it is the gap the principles were intended to narrow.
The paper also commits specific regulators to specific work. It records that the MHRA would develop guidance on the transparency and interpretability of AI as a medical device, working with the Information Commissioner's Office and the National Data Guardian on data protection and patient trust.
Non-statutory principles depend for their effect on how each regulator chooses to apply them. The framework does not bind a regulator that declines to prioritise them.
What has happened since
In February 2025 the government said that most AI systems should be regulated at the point of use, and that existing expert regulators are best placed to do this. That continues the 2023 approach rather than replacing it. The Labour government signalled in its 2024 election manifesto and in the 2024 King's Speech that it would introduce binding regulation on the handful of companies developing the most powerful AI models.
The House of Commons Library research briefing on AI regulation in the UK, published on 14 October 2026, records that the UK still has no AI-specific regulation or legislation covering AI as a technology. It notes that non-statutory principles are used alongside targeted legislation, and gives the Online Safety Act 2023 as an example: that Act empowers Ofcom to regulate online services, including AI chatbots, because it addresses a context rather than a technique.
The same briefing records that legislation on the most powerful models has not yet been forthcoming. It also notes that published AI policies have mainly focused on boosting the sector and promoting innovation in, and adoption of, AI.
One institutional commitment has been made. The AI Security Institute describes its mission as equipping governments with a scientific understanding of the risks posed by advanced AI, conducting research and developing and testing mitigations. That is a research and evaluation capability, not a regulatory power.
What is not yet known
What is not yet known is whether the commitment to legislate on frontier models will be honoured and on what timetable. The manifesto commitment exists; the Library briefing, written three years after the white paper, records that the legislation has not arrived. Until a bill is introduced and its scope defined, the framework in force is the one built in 2023 from existing regulators and non-statutory principles.
A second question is not settled by any document reviewed here. Whether a regulator required to have regard to cross-sectoral principles behaves differently in practice from one that is not is an empirical matter. No published evaluation of the framework's effect on regulator decisions was located.
Summary
The UK's framework for AI is a deliberate choice to regulate by context rather than by technology, and its central legislative commitment remains unfulfilled.
- The design is sector-based. Existing regulators apply the rules for the activity in which AI is used, rather than a single AI authority.
- The principles are non-statutory. Five cross-sectoral principles, set out in March 2023, guide regulator responses without binding them.
- The compliance problem was acknowledged. The white paper states that it is difficult for companies to identify which rules apply to a given use of AI.
- The legislative commitment is outstanding. Binding regulation for the most powerful models was signalled in 2024 and had not been introduced as of October 2026.
- Only a research body has been created. The AI Security Institute is tasked with understanding risk, not with regulating it.
Whether the government legislates on frontier models, and when, is not yet known.
Research direction and editorial oversight: Juno Uni Hera. AI-assisted research and drafting; human review is ongoing.
Sources
- Department for Science, Innovation and Technology. AI regulation: a pro-innovation approach (white paper). GOV.UK (2023). https://www.gov.uk/government/publications/ai-regulation-a-pro-innovation-approach/white-paper
- House of Commons Library. AI regulation in the UK. Research Briefing (2026). https://commonslibrary.parliament.uk/research-briefings/cbp-10003
- AI Security Institute. GOV.UK. https://www.gov.uk/government/organisations/ai-security-institute