AI Guide · Updated 14 November 2026
Risks and rules
AI law in the UK and EU
By drayz.ai · Updated · 14 min read

The UK and the European Union have taken very different routes to regulating artificial intelligence. The EU has passed the world's first comprehensive AI law, the AI Act, and began enforcing its rules for the most powerful AI models in August 2026. The UK has no single AI law. Instead it relies on existing laws, such as data protection, equality, consumer and online safety law, applied by existing regulators, plus a growing list of targeted new offences. This chapter explains both systems in plain English as they stand in November 2026, and what they mean for UK businesses. It is general information, not legal advice; for decisions that matter, speak to a qualified solicitor.
The short version
- UK: no AI Act. AI is regulated through the laws that already apply to what you do with it: UK GDPR and the Data (Use and Access) Act 2025, the Equality Act 2010, consumer protection law, the Online Safety Act 2023, copyright law, employment law and sector rules for finance, health and other industries. A dedicated bill for the most powerful AI models has been promised since 2024 but had not appeared by late 2026.
- EU: the AI Act sorts AI systems by risk. Some uses are banned outright, "high-risk" uses face strict requirements, chatbots and deepfakes must be labelled, and the makers of general-purpose AI models have their own duties. Rules for high-risk systems were delayed in 2026 to December 2027 and August 2028.
- UK businesses selling into the EU can be caught by the AI Act, just as they are by the EU GDPR.
The UK approach: principles, not a single law
In March 2023 the UK government published a white paper setting out a "pro-innovation" approach. Rather than creating a new AI regulator, it asked existing regulators to apply five cross-cutting principles within their own areas:
- Safety, security and robustness
- Appropriate transparency and explainability
- Fairness
- Accountability and governance
- Contestability and redress
This means the rules that apply to an AI system depend on what it is used for. A bank using AI to assess loans answers to the Financial Conduct Authority and the Information Commissioner's Office (ICO); a hospital using AI to read scans answers to the Medicines and Healthcare products Regulatory Agency (MHRA) and the Care Quality Commission; an online platform hosting AI-generated content answers to Ofcom.
Key regulators for AI in the UK include:
- ICO: data protection, including automated decision-making and AI training on personal data.
- Competition and Markets Authority (CMA): competition in AI markets and consumer protection, including fake reviews.
- Ofcom: online safety, including AI chatbots and AI-generated content on platforms.
- Financial Conduct Authority (FCA): AI in financial services.
- MHRA: AI used as a medical device.
- Equality and Human Rights Commission: discrimination.
The AI Security Institute (renamed from the AI Safety Institute in February 2025) is not a regulator. It is a government research body that tests advanced AI models for risks such as cyber attacks and biological misuse, usually with the cooperation of AI companies before release.
The missing UK AI bill
In its July 2024 King's Speech, the government said it would "seek to establish the appropriate legislation" for those developing the most powerful AI models. That bill was repeatedly delayed. The King's Speech of 13 May 2026 contained 37 bills but no comprehensive AI bill. Instead it included:
- A Regulating for Growth Bill, which would put regulatory "sandboxes" on a legal footing, allowing rules to be temporarily relaxed so that new products, including AI, can be tested in the real world.
- A Police Reform Bill, which would create a legal framework for police use of facial recognition and similar technology, with an independent regulator.
Committees in Parliament, including the Joint Committee on Human Rights, have called for a dedicated AI law, and private members' bills on AI have been introduced, but without government support they are unlikely to become law. For now, the UK's approach remains voluntary commitments from frontier AI companies plus existing law.
UK data protection and AI
For most UK organisations, data protection law is the most important AI law. If an AI system uses personal data, whether to train a model, to make decisions about people or simply because staff paste customer details into a chatbot, UK GDPR applies.
The Data (Use and Access) Act 2025 received Royal Assent on 19 June 2025, and most of its data protection changes came into force on 5 February 2026. The most important change for AI concerns automated decision-making:
- Previously, UK GDPR largely prohibited decisions with legal or similarly significant effects made solely by automated means, unless one of a few narrow conditions applied.
- Under the new rules, organisations can generally make such decisions on any lawful basis, including legitimate interests, provided safeguards are in place: people must be told about the decision, be able to make representations, obtain human intervention and contest the outcome.
- Stricter limits still apply where significant decisions rely on special category data, such as health, ethnicity or religion.
This makes the UK noticeably more permissive than the EU for automated decisions in areas such as recruitment and lending, but it does not remove the need for fairness, transparency and accuracy. The ICO publishes guidance on AI and data protection and has been preparing a statutory code of practice on AI and automated decision-making.
Practical points for businesses using AI with personal data:
- Know what data goes into AI tools and where it is processed and stored.
- Check supplier contracts and whether your data is used to train their models.
- Carry out a data protection impact assessment for higher-risk uses.
- Tell people when AI is used to make significant decisions about them, and give them a route to a human.
Discrimination, employment and consumer law
Equality Act 2010. If an AI system used in hiring, lending, insurance, housing or services produces discriminatory outcomes based on protected characteristics such as race, sex, age or disability, the organisation using it can be liable, even if it bought the system from someone else and did not intend the result.
Employment. Using AI to screen CVs, monitor staff or allocate work raises issues under data protection, equality and employment law. Transparency with staff and candidates, and human review of important decisions, are strongly advisable.
Consumer protection. The Digital Markets, Competition and Consumers Act 2024 gave the CMA power to fine businesses directly for consumer law breaches, and from April 2025 banned fake reviews, including reviews generated by AI and presented as genuine. Misleading claims about what an AI product can do can also breach consumer law.
Online safety, deepfakes and criminal law
The UK has created several targeted offences aimed at harmful uses of AI:
- The Online Safety Act 2023 requires platforms, search services and many AI chatbot services to protect users, especially children, from illegal and harmful content. Ofcom has confirmed that generative AI chatbots that let users share content, and AI search tools, can fall within the Act.
- Sharing intimate images without consent, including deepfakes, is a criminal offence, and since 6 February 2026, section 138 of the Data (Use and Access) Act 2025 has made it an offence to create, or ask someone else to create, a sexually explicit deepfake of an adult without consent, even if it is never shared.
- The Crime and Policing Act 2026, which received Royal Assent on 29 April 2026, created new offences of making, adapting or supplying AI models optimised to generate child sexual abuse material, and of supplying tools used to create purported intimate images. Companies as well as individuals can be prosecuted. It also lets ministers bring illegal AI-generated content and AI services more firmly within the Online Safety Act.
Copyright and AI in the UK
Copyright has been the most hotly contested AI issue in the UK. AI developers want to train models on large amounts of text, images and music; creators want permission and payment.
- Under the Copyright, Designs and Patents Act 1988, the UK's text and data mining exception (section 29A) only covers non-commercial research. Commercial AI training on copyrighted works generally needs a licence.
- In December 2024 the government consulted on a new exception allowing commercial AI training unless rights holders opted out. This prompted a major campaign by musicians, authors, publishers and other creators, and long battles in the House of Lords during the passage of the Data (Use and Access) Bill.
- On 18 March 2026, the government published its report and impact assessment on copyright and AI, as required by that Act. It confirmed that the opt-out exception was no longer its preferred option and that there would be no immediate change to copyright law. Instead it will gather more evidence, watch court cases and the licensing market, and work on transparency, labelling of AI content and protections against digital replicas.
- The UK is unusual in protecting "computer-generated works" with no human author. The government has signalled that this protection is likely to be removed, but deferred a final decision.
In practice, UK businesses using AI tools should check their suppliers' terms on copyright, avoid using AI to imitate specific artists or brands, and remember that AI outputs can still infringe existing works.
The EU AI Act: how it works
The EU Artificial Intelligence Act (Regulation (EU) 2024/1689) entered into force on 1 August 2024. It applies to anyone who places AI systems on the EU market or whose AI outputs are used in the EU, wherever they are based. It sorts AI into risk levels:
1. Prohibited AI practices
Since 2 February 2025, certain uses have been banned in the EU, including:
- Manipulative or deceptive AI that causes significant harm.
- Exploiting people's vulnerabilities due to age, disability or social situation.
- "Social scoring" of people leading to unjustified detrimental treatment.
- Predicting crime based solely on profiling or personality traits.
- Building facial recognition databases by untargeted scraping of images from the internet or CCTV.
- Emotion recognition in workplaces and schools, except for medical or safety reasons.
- Biometric categorisation to infer sensitive characteristics such as race, religion or sexual orientation.
- Real-time remote biometric identification in public spaces by police, except in narrowly defined cases.
From 2 December 2026, the 2026 amendments add bans on AI systems designed to generate non-consensual intimate images of identifiable people ("nudifier" apps) or child sexual abuse material.
2. High-risk AI systems
AI used in sensitive areas is classed as high-risk and must meet strict requirements: risk management, data quality, technical documentation, record-keeping, transparency, human oversight, accuracy, robustness and cybersecurity, plus registration in an EU database. High-risk areas include:
- Biometric identification.
- Critical infrastructure.
- Education and exams.
- Employment, recruitment and management of workers.
- Access to essential services, such as credit scoring and insurance pricing.
- Law enforcement, migration and border control.
- Justice and democratic processes.
- AI used as a safety component in regulated products such as medical devices, machinery and toys.
3. Transparency duties
From 2 August 2026, people must be told when they are talking to an AI system such as a chatbot, unless it is obvious, and deepfakes and certain AI-generated text published to inform the public must be labelled. Providers of generative AI must mark synthetic content in a machine-readable way; systems already on the market before August 2026 have until 2 December 2026 to do so.
4. Minimal-risk AI
Most AI, such as spam filters, recommendation engines or AI in video games, faces no specific new obligations, although all organisations using AI were initially required to ensure their staff had sufficient AI literacy.
General-purpose AI models
The AI Act has a separate chapter for general-purpose AI (GPAI) models, such as the large language models behind ChatGPT, Claude, Gemini, Llama and Mistral.
- Since 2 August 2025, providers of new GPAI models must keep technical documentation, give information to businesses building on their models, have a policy to comply with EU copyright law (including respecting opt-outs from text and data mining), and publish a summary of the content used for training.
- Models with "systemic risk", broadly the most powerful models, trained using more than 10^25 floating-point operations of computing or designated by the Commission, must also evaluate and mitigate risks, report serious incidents and ensure strong cybersecurity.
- A voluntary General-Purpose AI Code of Practice, published on 10 July 2025, gives providers a way to show compliance. Many major AI developers signed it, although Meta declined and some companies signed only parts of it.
- From 2 August 2026, the European Commission's AI Office gained full enforcement powers over GPAI providers: it can demand documentation, evaluate models, order corrective measures, restrict or withdraw models from the EU market and impose fines.
The 2026 changes: the AI Omnibus
By late 2025 it was clear that technical standards and national regulators would not be ready in time for the high-risk rules due in August 2026. In November 2025 the Commission proposed a "Digital Omnibus" to simplify EU digital rules. After a provisional deal on 7 May 2026, the AI Omnibus (Regulation (EU) 2026/1744) was published on 24 July 2026 and entered into force on 27 July 2026. Its main changes:
| Rule | Original date | New date |
|---|---|---|
| High-risk AI in areas such as employment, education, credit and law enforcement (Annex III) | 2 August 2026 | 2 December 2027 |
| High-risk AI as a safety component of regulated products (Annex I) | 2 August 2027 | 2 August 2028 |
| Machine-readable labelling for generative AI already on the market | 2 August 2026 | 2 December 2026 |
| Ban on nudifier and child abuse image generators | Not in original Act | 2 December 2026 |
Other changes included extending lighter-touch rules for small businesses to "small mid-caps" with up to around 500 staff, an EU-level regulatory sandbox, turning the AI literacy duty into something the Commission and member states must promote rather than a strict obligation on every business, and allowing limited use of sensitive personal data to detect and correct bias. Rules for general-purpose AI models and transparency for new systems were not delayed.
Penalties under the AI Act
Fines are set as a fixed amount or a percentage of worldwide annual turnover, whichever is higher (with lower caps for small businesses):
- Up to €35 million or 7% for using prohibited AI practices.
- Up to €15 million or 3% for breaching most other obligations, including those for general-purpose AI models.
- Up to €7.5 million or 1% for supplying incorrect or misleading information to authorities.
What UK businesses should do now
Whether you build AI or simply use it, a sensible checklist for 2026-2027:
- List your AI uses. Include tools staff use informally, such as chatbots and transcription apps.
- Check personal data. Identify where personal data goes, update privacy notices, and run impact assessments for higher-risk uses.
- Keep a human in the loop for decisions that significantly affect people, and give them a way to challenge outcomes.
- Test for bias in any AI used for hiring, lending, pricing or access to services.
- Label AI interactions. Tell customers when they are talking to an AI agent; this is required in the EU and good practice in the UK.
- Check supplier contracts for data use, security, copyright indemnities and where processing happens.
- Work out your EU exposure. If you have EU customers or your AI outputs are used in the EU, plan for the AI Act, especially if your use could be high-risk by December 2027.
- Train your staff in safe and effective AI use.
- Watch for changes. UK copyright policy, a possible frontier AI bill, the ICO's AI code and EU guidance are all still developing.
Frequently asked questions
Does the UK have an AI Act?
No. As of November 2026 the UK has no single AI law. AI is regulated through existing laws such as UK GDPR, the Data (Use and Access) Act 2025, the Equality Act 2010, consumer law and the Online Safety Act, plus new targeted criminal offences. A bill for the most powerful AI models has been promised since 2024 but was not included in the May 2026 King's Speech.
Does the EU AI Act apply to UK companies?
It can. The AI Act applies to providers who place AI systems or models on the EU market and to providers and users outside the EU whose AI outputs are used in the EU. A UK business selling AI products or services to EU customers should check whether it is covered.
When do the EU high-risk AI rules apply?
After the 2026 AI Omnibus amendments, most high-risk rules, such as those for AI in recruitment, education and credit scoring, apply from 2 December 2027. Rules for AI built into regulated products such as medical devices and machinery apply from 2 August 2028.
Do I have to tell customers they are talking to an AI?
In the EU, yes: since 2 August 2026, people must be told they are interacting with an AI system unless it is obvious. In the UK there is no general AI disclosure law, but data protection, consumer and sector rules often require transparency, and it is good practice.
Can AI companies train on copyrighted material in the UK?
Commercial training on copyrighted works generally requires permission, because the UK's text and data mining exception only covers non-commercial research. In March 2026 the government dropped its preferred plan for a broader exception with an opt-out and said there would be no immediate change to the law.
What are the fines under the EU AI Act?
Up to €35 million or 7% of worldwide annual turnover for prohibited practices, up to €15 million or 3% for most other breaches including general-purpose AI obligations, and up to €7.5 million or 1% for misleading information to regulators.
Key takeaways
- The UK regulates AI through existing laws and regulators guided by five principles; there is still no UK AI Act.
- The Data (Use and Access) Act 2025 loosened UK rules on automated decision-making from February 2026, with safeguards.
- New UK offences target AI-generated child abuse material and sexually explicit deepfakes.
- The UK dropped its plan for an opt-out copyright exception for AI training in March 2026.
- The EU AI Act bans certain uses, strictly regulates high-risk AI and imposes duties on general-purpose AI models; enforcement of the GPAI rules began on 2 August 2026.
- The 2026 AI Omnibus delayed high-risk rules to December 2027 and August 2028 and added bans on nudifier apps.
- UK businesses with EU customers may need to comply with the AI Act.
Previous chapter: AI and jobs in the UK · Back to the complete guide to artificial intelligence