EU AI Act transparency rules: what UK businesses should do now

What changed

The European Union's transparency rules for AI take effect on 2 August 2026. Set out in Article 50 of the EU AI Act, they require organisations to tell people when they are interacting with an AI system, and to mark AI-generated images, audio, video and text so it can be detected as artificial. The European Commission's consultation on its draft transparency guidelines closed on 3 June 2026, and a voluntary Code of Practice on marking AI-generated content is expected to be finalised in June 2026. Under the Digital Omnibus agreed on 7 May 2026, one part of the rules has a later deadline: the machine-readable marking requirement falls on 2 December 2026 for generative AI systems that were already on the market.

Why it matters for UK business

The first question for a UK firm is whether these rules apply to it at all, and the honest answer is that they often do, but not always. The AI Act reaches beyond the EU's borders. It applies to providers and deployers based outside the Union, the UK included, where the output of their AI system is used inside the EU. A UK business that runs an AI chatbot used by customers in the EU, or that publishes AI-generated content seen by EU audiences, is likely in scope. A firm whose AI use is wholly domestic is generally not directly bound, though it inherits the same expectations indirectly, because the AI tools it buys will increasingly mark and disclose by default to serve their EU customers.

The obligations themselves are not onerous, but they are specific, and they split across two roles. Providers, the people who build or supply the system, must design AI that interacts with people so users know they are dealing with a machine, and must mark synthetic output in a machine-readable form. Deployers, the people who use it, must disclose deep fakes, label AI-generated text published to inform the public on matters of public interest, and tell people when emotion-recognition or biometric-categorisation systems are in use. There are sensible exceptions: disclosure is not required where the AI involvement is obvious, where content has had genuine human editorial review, or where the output is evidently artistic or satirical.

For most UK SMEs the practical exposure is narrower than the headlines suggest, and the sensible posture is preparation rather than alarm.

What to do, and what not to do

Do:

• Work out your role and your reach first. Decide whether you are a provider or a deployer of each AI system you use, and whether any output reaches the EU. That single assessment tells you what, if anything, applies.
• Add a clear AI-interaction notice to customer-facing chatbots and assistants. Telling users they are talking to AI is low-cost, is good practice under UK GDPR transparency principles anyway, and aligns you with the direction of travel.
• Keep an inventory of where you generate AI content and where you label it. The marking and disclosure duties are easier to meet when you already know which workflows produce synthetic text, images or audio.
• Check what your vendors are doing. Much of the marking obligation sits with the model providers; confirm that the tools you rely on will support machine-readable marking before the December 2026 date.

Do not:

• Assume Brexit puts you out of scope. The test is where the output is used, not where your company is registered.
• Treat this as a one-off compliance project. Transparency is a standing operating practice: disclosure, labelling and a human-review record, maintained as workflows change.
• Confuse the EU rules with current UK law. The UK has not adopted the AI Act and is pursuing a lighter, regulator-led approach. The reason to act is EU market access and good practice, not a UK statutory duty.
• Wait for the enforcement detail before doing the cheap, sensible things. Disclosure notices, a content inventory and vendor checks are worth doing on their own merits.

Where The AI Consultancy fits

Mapping which of your AI systems are in scope, and turning Article 50 into a short, practical checklist of disclosures and labels, is the kind of work our AI Readiness assessments cover. If you sell into the EU or publish AI-assisted content to EU audiences and want a clear read on what applies before the August deadline, that is a scoped piece of analysis, not a compliance overhaul. Our guides to the EU AI Act for UK SMEs and the UK AI compliance checklist set out the wider picture.

Dates and obligations in this briefing are drawn from the published text of Article 50 of the EU AI Act, the European Commission's transparency consultation and Code of Practice timeline, and the Digital Omnibus agreement of 7 May 2026, as at 6 June 2026. This briefing is general information, not legal advice; confirm your specific obligations with a qualified adviser.