The situation
The European Commission published guidance on 20 July explaining how Article 50 transparency obligations under the EU AI Act apply from 2 August 2026. The guidance does not create new duties; it clarifies scope, exemptions and compliance expectations for providers and professional deployers of specified AI systems.
Covered providers of directly interactive AI systems must tell users they are interacting with AI unless this is obvious. Providers of eligible generative systems must apply effective, reliable, robust and interoperable machine-readable marking to AI-generated or manipulated output. Professional deployers must disclose the use of deepfakes and certain emotion-recognition or biometric-categorisation systems, and clearly label specified AI-generated public-interest text where it has not received substantive human review or editorial control. For systems placed on the market before 2 August, the marking and detection requirement applies from 2 December 2026; content generated before 2 August does not require retroactive labelling.
Why it matters
The guidance gives firms and national authorities a common operational reference less than two weeks before the obligations begin. It also confirms that certain providers outside the EU may be covered where their AI-system output is used in the Union, extending the compliance significance beyond EU-based companies.
Enforcement will mainly rest with national market-surveillance authorities, while the EU AI Office has defined responsibilities in certain general-purpose-AI and designated large-platform or search-engine cases. Non-compliance can carry fines of up to €15 million or 3% of worldwide annual turnover, subject to proportionality considerations for smaller firms. The Commission’s voluntary transparency code is one possible compliance route, but organisations may use adequately equivalent alternatives.
What changed
On 20 July 2026, the Commission published interpretive and implementation guidance for Article 50, including coverage, exemptions, compliance routes, enforcement roles and the transition for certain pre-existing systems.
Impacts
Economic Pressure
The guidance raises near-term compliance demands for covered AI providers and professional deployers before 2 August, with maximum penalties of €15 million or 3% of prior-year worldwide turnover for non-compliance.
Diplomacy
The rules can apply to non-EU providers when their AI-system output is used in the EU, giving the framework cross-border regulatory reach while primary enforcement remains national.
Political Stability
The Commission has clarified EU-wide transparency expectations, but practical consistency will depend on how national market-surveillance authorities enforce the rules after they begin applying.
What to watch
- Do national market-surveillance authorities publish coordinated enforcement priorities or implementation guidance before the Article 50 obligations begin on 2 August 2026?
- Do major AI providers and professional deployers adopt machine-readable marking and disclosure processes that meet the Commission’s stated requirements by the start date?
Evidence and uncertainty
Low confidence. Low confidence. 3 retained references from 1 reporting origins (European Commission). Claim-level independent corroboration has not been established. Confidence concerns the evidence for the reported development, not the probability of an outlook.
Useful context
Article 50
The provision of the EU AI Act setting transparency obligations for certain AI-system providers and deployers.
Its requirements on notices, content marking and disclosures begin applying on 2 August 2026.
EU AI Office
The Commission body responsible for specified functions in the implementation and supervision of the EU AI Act.
It has jurisdiction in certain general-purpose-AI and designated large-platform or search-engine cases.
Geographic context
- 01Brussels, Belgium (city)