Brexit did not put UK companies outside the EU AI Act. The Act follows the market, not the head office, so a firm in Manchester with customers in Munich can be bound by it. The question is which of your systems are caught, and by when.
The short answer
- The Act applies by market. A UK provider or deployer whose AI system, or whose AI output, is placed on or used in the EU is in scope wherever it is based.
- The 2 August 2026 date still stands for the general application of the Act and for transparency duties on chatbots and synthetic content.
- The 2025 Digital Omnibus, in force since July 2026, pushed the heaviest high-risk obligations back to December 2027, and product-embedded ones to August 2028. The delay is real, and it is not a reprieve.
- Penalties reach the higher of a fixed cap or a percentage of worldwide turnover, so scope is a board matter, not a compliance footnote.
- The first task is classification: list the AI systems that touch the EU and place each in the Act's risk tiers.
Why a UK company is in scope at all
Article 2 sets the reach of the Act, and it is deliberately wide. It binds providers that place AI systems on the EU market or put them into service there, deployers established or located in the EU, and, critically for UK firms, providers and deployers outside the EU where the output produced by the system is used in the EU. A UK software company selling an AI feature to European customers is a provider on the EU market. A UK group whose model scores applications for an EU subsidiary is producing output used in the EU. Neither has an EU office, and both are in scope.
This mirrors the reach UK boards already learned from the GDPR, so the mechanism is familiar even where the detail is new. The mistake to avoid is the assumption that a UK base is a shield. It is not, and the enforcement powers behind the Act make the point worth checking rather than assuming.
The dates, after the delay
The Act entered into force in August 2024 and phases in over several years. In November 2025 the European Commission proposed a Digital Omnibus that deferred the high-risk obligations, and it entered into force in July 2026, so the revised timeline is now the law rather than a proposal.
| Date | What applies | Who should care |
|---|---|---|
| February 2025 | Prohibited practices banned, such as social scoring and certain biometric uses. | Anyone operating these; they are off the table. |
| August 2025 | Obligations on general-purpose AI models begin. | Providers of foundation models, and firms building on them. |
| 2 August 2026 | General application of the Act, and Article 50 transparency duties on chatbots and AI-generated content. Enforcement powers over general-purpose AI switch on. | Almost every company using a customer-facing chatbot or publishing synthetic media into the EU. |
| 2 December 2027 | High-risk obligations for stand-alone Annex III systems, deferred from August 2026 by the Omnibus. | Users of AI in hiring, credit, biometrics, education and similar. |
| 2 August 2028 | High-risk obligations for AI embedded in regulated products under Annex I. | Manufacturers of regulated products with AI inside them. |
The delay changes the sequence of work rather than its size. Transparency duties are live on the 2026 date, so anything conversational or generative facing EU users needs handling now. The high-risk build, which is the expensive part, has more runway, and using that runway to prepare properly is the point of the extension.
The transparency duties that bite first
Article 50 is the part most UK companies underestimate, because it applies broadly rather than only to high-risk systems. People must be told when they are interacting with an AI system unless it is obvious. AI-generated or manipulated content, including synthetic audio, images and video, must be marked as such in a machine-readable way. A customer-service chatbot, a marketing tool that produces synthetic imagery, or a voice agent handling EU calls all fall here, and all reach the standard on the 2026 date. This is the cheapest obligation to meet and the easiest to miss, because it sits in marketing and service functions rather than in a risk team.
What a board should do before the next EU-facing launch
The Act rewards early classification and punishes late discovery. Three steps put a board ahead of it.
Build the inventory around the EU nexus
List the AI systems the organisation provides or deploys, and mark which ones touch the EU market or produce output used in the EU. That single column decides scope. A system with no EU nexus is a UK regulatory matter under existing law, which is covered in the UK's sector-regulator approach. A system with an EU nexus is also an EU AI Act matter.
Classify each EU-facing system into a risk tier
Prohibited, high-risk, limited-risk with transparency duties, or minimal. Most enterprise systems land in limited-risk, where Article 50 transparency is the obligation. The ones that land in high-risk carry the heavier duties on risk management, data governance, human oversight and documentation, on the 2027 timeline.
Assign the transparency work now
Because the 2026 date is close and the duty is operational, name an owner for chatbot disclosure and synthetic-content marking before the next campaign or release, rather than treating the whole Act as a 2027 programme.
The constraint
The binding limit for most UK firms is an accurate map of which systems have an EU nexus, rather than the depth of any single compliance control. Investing in a full high-risk conformity programme before that map exists spends money on systems that may be out of scope while missing the transparency duty that is already live, because the deciding question is where each system touches the EU market, and only the inventory answers it.
The question that settles exposure
For each AI system, ask whether it is placed on the EU market, deployed by an EU entity, or produces output used in the EU, and if so, which risk tier it falls into. A company that can answer that for its estate knows its real exposure and its real deadlines. One that cannot is guessing, and the Act is not a good thing to guess about, given that the penalties are set against worldwide turnover.
Sources
- Regulation (EU) 2024/1689 (the EU AI Act), in force from 1 August 2024, Articles 2, 50 and 99.
- European Commission, Regulatory framework on AI.
- European Commission Digital Omnibus on AI, published 19 November 2025, in force from 27 July 2026, deferring high-risk obligations for Annex III systems to 2 December 2027 and Annex I to 2 August 2028.