The EU's New AI Disclosure Law Skips UK Classrooms Entirely

A new rule most UK teachers haven't heard of

On 2 August, the European Commission began enforcing Article 50 of the EU AI Act: a transparency duty requiring that any AI system interacting directly with people identify itself as AI, clearly, before or at the very start of the conversation. For chatbots aimed at children, the disclosure has to be child-friendly and age-appropriate, not buried in a terms-of-service page nobody reads.

The European Commission's own guidance is blunt about what counts as compliant: "a statement buried in terms and conditions, documentation or a separate URL is insufficient on its own." The AI has to say, in effect, I am not a person, somewhere a child will actually see it.

This is a different, narrower rule from the AI Act's high-risk classification for education systems (admissions, grading, behavioural monitoring), which was pushed back to December 2027 under the summer's Digital Omnibus amendment. Disclosure went ahead on schedule. Classification didn't. That distinction matters, because it means the part of the law that touches ordinary classroom chatbots and AI tutors is already live, while the part that would formally regulate exam-adjacent AI tools is not.

Why this reaches further than Brussels

Most of the AI tools UK teachers and pupils use daily — homework helpers, AI tutoring companions, essay-feedback bots, revision chatbots — are built by companies that also sell into the EU. Few vendors maintain separate product builds for separate jurisdictions, so a disclosure requirement written for European users often ends up shipping everywhere, UK included, as a side effect of compliance engineering rather than choice.

But that's an assumption, not a guarantee. Article 50 does not bind schools in England, Scotland, Wales or Northern Ireland, because the UK left the EU regulatory orbit and has no equivalent statute. A vendor can quietly comply for its EU users and leave the UK version of the same product unchanged. Nothing forces a British school's AI reading assistant, homework bot or maths tutor to tell an 11-year-old it isn't human.

The UK's closest domestic protections sit elsewhere and cover different ground. The Information Commissioner's Office's Age Appropriate Design Code sets standards for how services handle children's data and requires clear terms, but it was written before generative AI chatbots existed and doesn't mandate an explicit "I am an AI" disclosure at the start of a conversation. Keeping Children Safe in Education's 2026 update addresses safeguarding risks from AI, not whether a chatbot must announce its own nature. There is, at present, no UK rule that does what Article 50 now does across the Channel.

A pupil in Dublin using an AI homework helper is now legally guaranteed a plain statement that they're talking to a machine. A pupil in Belfast, using the same product, may not be.

The content-marking half of the rule

Article 50 also requires AI-generated or manipulated content, including deepfake-style images, video and audio, to carry machine-readable and human-readable marks so it can be identified as synthetic. Minor edits like grammar correction are exempt, but AI-generated material on matters of public interest is not. For schools, this has a quieter but real relevance: AI-generated video explainers, AI avatars used in flipped-classroom content, or AI-written material passed off as staff-authored communications to parents would need labelling under EU rules. UK schools producing similar material face no comparable statutory requirement, though several exam boards' academic integrity guidance already asks staff to be transparent about AI use in materials they produce.

What to do

• Ask your platform providers directly whether pupil-facing chatbots disclose their AI nature at the start of a session, and whether that disclosure is UK-visible or EU-only. Don't assume parity.
• If a tool already shows an EU-style disclosure to your students, treat it as a floor, not a policy. Build your own AI-use disclosure expectations into acceptable-use agreements rather than relying on vendor defaults.
• Where staff use AI to generate parent-facing content, such as newsletters, video summaries or draft reports, disclose it voluntarily. It costs nothing and pre-empts the trust problem the EU rule is designed to solve.
• Flag the gap to whoever owns AI policy in your trust or local authority. Regulatory divergence from the EU is a known, growing feature of UK edtech procurement, not a one-off.

What to watch

Watch whether the Department for Education or the ICO moves to close this gap before the Annex III high-risk deadline lands in December 2027. Watch too whether major AI vendors quietly extend EU-style disclosure to UK users anyway, since maintaining two product behaviours is expensive and most will default to the stricter one. If they don't, UK schools will be running the same generative AI tools as their European counterparts under a materially weaker transparency standard, with no domestic law requiring anything different.