The United Kingdom's AI strategy rests on three documents: a March 2023 white paper that left AI rules to existing regulators, the AI Opportunities Action Plan of January 13, 2025, which set 50 recommendations on computing power, data, and adoption, and a rebranded AI Security Institute that tests advanced models. Government statistics show business use rising from around 12 percent of firms with 10 or more employees in September 2023 to around 35 percent in June 2026, though most firms use only one or two AI tools.
Key takeaways
- The 2023 white paper chose five principles applied by existing regulators over a new AI regulator or a new AI statute.
- The Action Plan has 50 recommendations; the government agreed with 47, partly agreed with two, and reported meeting commitments on 38 of the 50 actions by January 29, 2026.
- The AI Security Institute was renamed on February 14, 2025 and now concentrates on security risks such as cyber-attacks and chemical or biological misuse.
- The Office for National Statistics (ONS) reports about 35 percent of UK firms with 10 or more employees using at least one AI technology in June 2026.
- The ONS labels its source data "official statistics in development," so the figures carry extra uncertainty.
Rules first: the 2023 white paper
The government presented "AI regulation: a pro-innovation approach" to Parliament on March 29, 2023. It named five cross-sector principles: safety, security and robustness; appropriate transparency and explainability; fairness; accountability and governance; and contestability and redress. The paper proposed that the government issue these principles on a non-statutory basis and that existing regulators apply them within their own areas, such as health, finance, and competition. It stated that the government did not intend to create a new AI regulator. A later legal duty on regulators to have due regard to the principles was floated only for a time when parliamentary time allows and monitoring shows the voluntary approach is failing.
That design differs from the European Union's single AI Act. A UK developer or deployer deals with several regulators, each reading the principles through its own sector rules.
What the Action Plan promised
The AI Opportunities Action Plan was published on January 13, 2025 by the Department for Science, Innovation and Technology (DSIT). In the foreword, Secretary of State Peter Kyle says he commissioned Matt Clifford to devise it. Its 50 recommendations fall into three sections, and several carry numbers that can be checked:
- Recommendation 2 calls for at least a 20-fold expansion of the public AI Research Resource by 2030.
- Recommendation 4 proposes AI Growth Zones to speed up data center construction, with the Culham site of the UK Atomic Energy Authority suggested as a pilot site.
- Recommendation 7 asks the government to identify at least five high-impact public datasets for AI researchers.
In its response of the same date, the government marked 47 recommendations as agreed and two (13 and 21) as partly agreed, and it described a first growth zone at Culham starting at 100 megawatts with plans to reach 500 megawatts. A "one year on" report dated January 29, 2026 states that the government had met its commitments on 38 of the 50 actions. The report is a self-assessment, not an independent audit.
The AI Security Institute
On February 14, 2025, Technology Secretary Peter Kyle announced at the Munich Security Conference that the AI Safety Institute would become the AI Security Institute. The government said the institute's work would continue with a sharper focus on serious risks with security implications, including AI-enabled chemical and biological weapons, cyber-attacks, fraud, and child sexual abuse, and that it would not focus on bias or freedom of speech. A criminal misuse team was to work with the Home Office.
The institute describes itself as a research organization inside DSIT. It tests leading AI systems before and after release, publishes open evaluation tools such as Inspect, and has awarded more than GBP 15 million in grants. Its pre-deployment access to models comes through collaboration with AI companies. The institute's own page does not claim statutory powers.
How widely firms use AI
On July 20, 2026, the ONS published "Artificial intelligence in UK businesses: 2023 to 2026," based on the Business Insights and Conditions Survey. The table collects its main figures for businesses with 10 or more employees unless noted.
| Measure | Figure |
|---|---|
| Firms using at least one AI technology, September 2023 | Around 12 percent |
| Firms using at least one AI technology, June 2026 | Around 35 percent |
| Businesses with 0 to 9 employees using AI, June 2026 | 28 percent |
| Businesses with 250 or more employees using AI, June 2026 | 49 percent |
| Average number of AI technologies per adopting firm | 1.4 in 2023, 1.6 in 2026 |
| Large language models, June 2026 | 18 percent of firms |
| Visual content creation, June 2026 | 16 percent of firms |
The ONS reads the pattern as wider adoption without much deeper use. Improving business operations was the most reported purpose in every size band. The survey measures whether a firm reports using a technology, not how much work the technology does, and it covers firms rather than individual workers.
Gaps in the public record
Several points could not be settled from primary sources for this article. Whether the UK Parliament has moved toward a dedicated AI bill, and whether the AI Security Institute might receive statutory testing powers, is reported differently by secondary outlets, so the status is left open here. Readers who need the current position can check the UK Parliament bill tracker and UK government publications. The Action Plan also does not say how its 50 actions were weighted, so "38 of 50" counts actions without ranking their importance.
This article gives general information about public policy. It is not legal advice, and organizations with duties under UK regulation should consult the relevant regulator or a qualified adviser.
FAQ
Does the UK have an AI Act like the European Union?
The 2023 white paper chose not to create one and relied on existing regulators applying five principles. Reports on a future bill differ, and this article could not confirm a current one from a primary source.
What is the difference between the AI Safety Institute and the AI Security Institute?
They are the same body. It was renamed on February 14, 2025, and the government said the new name reflects a focus on security risks. The government also said it would not work on bias or freedom of speech.
How many UK businesses use AI?
The ONS reports about 35 percent of firms with 10 or more employees in June 2026, with 28 percent among the smallest firms and 49 percent among those with 250 or more employees. Figures from different surveys use different samples and wording, so they should not be mixed.
Did the government deliver the Action Plan?
Its own January 2026 report says it met commitments on 38 of 50 actions. That is a self-reported count and does not measure outcomes.
Related reading
- AI and robotics: where the fields meet
- The main components of an AI system
- How much mathematics AI work needs
- AI-generated art explained
More in AI Basics.
Last verified: October 2026
Sources
- AI Opportunities Action Plan, Department for Science, Innovation and Technology, January 13, 2025
- AI Opportunities Action Plan: government response, Department for Science, Innovation and Technology, January 13, 2025
- AI Opportunities Action Plan: one year on, Department for Science, Innovation and Technology, January 29, 2026
- AI regulation: a pro-innovation approach (white paper), UK Government, March 29, 2023
- UK Government news release on the renaming of the AI Safety Institute, February 14, 2025
- About the AI Security Institute, AI Security Institute, 2026
- Artificial intelligence in UK businesses: 2023 to 2026, Office for National Statistics, July 20, 2026