EU AI Act Article 50: what changes on 2 August 2026
A practical evidence guide to the EU AI Act transparency obligations that begin applying on 2 August 2026, including interactive AI, AI-generated content, deepfakes and the difference between provider marking and deployer disclosure.
Last reviewed 2026-07-26 · Editorial methodology
What changes on 2 August 2026 under Article 50?
Article 50 transparency obligations begin applying on 2 August 2026. In-scope cases include informing people when they are interacting with certain AI systems, machine-readable marking of AI-generated or manipulated content by providers, and disclosures by deployers for deepfakes, certain public-interest text, emotion recognition and biometric categorisation. The exact obligation depends on the system, role and use case.
Does every business using AI have to label everything?
No. Article 50 is not a blanket label-everything rule. Different duties apply to providers and deployers, and the regulation contains scope conditions and exceptions. Organisations should map the systems and content they actually provide or deploy rather than treating every AI-assisted task as legally identical.
What should an organisation check before 2 August?
Start with an inventory of interactive AI and generative systems, identify who is provider or deployer for each use, document where generated or manipulated content reaches people, review notices and labelling, assign an accountable owner, and retain evidence of how the chosen controls meet the applicable requirement.
Why does this matter beyond legal compliance?
Transparency is also a trust and operating-design problem. A technically correct notice that users never see, or a label with no internal owner or audit trail, is weak governance. The useful question is whether a person can understand when AI is involved and whether the organisation can demonstrate how that disclosure or marking is produced.
What does the current evidence say?
| Claim | Source organisation | Date | Evidence |
|---|---|---|---|
| The European Commission published final Article 50 transparency guidelines on 20 July 2026 and states that the transparency obligations apply from 2 August 2026. | European Commission, AI Office | 2026-07-20 | Primary source |
| The European Commission says its enforcement powers for obligations on providers of general-purpose AI models enter into application from 2 August 2026. | European Commission, AI Office | 2026-05-01 | Primary source |
Limitations
This guide is a practical evidence map, not a universal statement that every UK organisation or every use of AI is in scope. Applicability depends on role, system, market, content and the relevant provisions and exceptions. Organisations operating in or serving the EU should map their actual circumstances against the current legal text and Commission guidance.
A counterpoint worth keeping
More labels do not automatically create more transparency. Repetitive or badly placed notices can become wallpaper. The stronger control is a proportionate disclosure that is visible at the right moment, backed by reliable technical marking where required and by records showing who owns the process.
Sources and provenance
- European Commission, AI Office · Guidelines on transparency obligations for providers and deployers of AI systems · 2026-07-20 · Primary source
- European Commission, AI Office · Quick Facts: Transparency rules for AI systems · 2026-07-20 · Primary source
- European Commission, AI Office · Guidelines for providers of general-purpose AI models · 2026-05-01 · Primary source
Continue the evidence trail
Not sure which book fits?
Use the rule-based book finder to narrow the 40-book catalogue by the problem you are actually trying to solve.
Find the right AI book