European Commission publishes Article 50 transparency guidance

The European Commission has published guidance setting out which providers and deployers must meet Article 50 transparency obligations under the EU AI Act.

Issued on 20 July 2026, the guidance explains duties that begin to apply on 2 August 2026: telling people when they are interacting directly with an AI system (for example chatbots and autonomous agents); applying machine‑readable marks to enable automated detection of AI‑generated or manipulated content; and disclosing use of emotion‑recognition or biometric categorisation systems. The Commission links the guidance to a Code of Practice and implementation Q&As aimed at deployers and market surveillance authorities.

The paper is framed as an interpretation of Article 50 rather than a change to the statute. The Commission says its aim is to help operators understand who counts as a "provider" and who counts as a "deployer" for the transparency duties, and to set out practical examples where the obligations apply — including scenarios common to HR teams, such as recruitment chatbots, candidate‑screening tools and surveillance systems that analyse behaviour or affect working conditions.

For HR leaders and talent‑tech vendors, the guidance contains three practical threads. First, where an individual is engaging directly with an automated system — a virtual recruiter, onboarding assistant or conversational agent — the system must make that interaction explicit. Second, systems that generate or materially manipulate content must carry machine‑readable marks that allow automated detection and tagging of AI output. Third, the Commission draws a distinction for higher‑sensitivity capabilities: employers and vendors must disclose if systems infer or use emotional state or biometric categories about a person.

Brussels frames the measures as targeted transparency requirements rather than an expansive new privacy regime. The Commission points to the Code of Practice and the Q&As to help national market surveillance authorities and deployers operationalise the rules, signalling that enforcement and technical implementation will be handled at member‑state level.

The guidance arrives as the EU moves from legislative text to operational rules for AI. Regulators across Europe have over the past year shifted from setting high‑level obligations to issuing interpretive material to ensure consistent application of the law. For the HR tech market — a sector that mixes cloud providers, platform vendors and employer deployers — the document seeks to reduce ambiguity about which party must place notices and machine marks, a recurring headache in multi‑party procurement chains.

What the Commission did not settle is every technical or enforcement detail. The guidance does not prescribe a single, binding format for the machine‑readable mark nor does it embed an EU‑wide technical specification for automated detection; those elements are left to follow‑up work by standardisers and market surveillance authorities. The paper also leaves open granular questions about liability in complex supply chains — for instance where a vendor supplies an embedded model to a platform that a separate employer deploys — and it does not set out how national authorities will prioritise checks or penalties in the early weeks of application.

For employers and HR teams, the immediate task is procedural: audit deployed systems to identify any that interact directly with people, generate content, or analyse emotion/biometrics; update candidate and employee notices; and clarify contractual responsibilities with vendors about marking and disclosure. Vendors will need to decide how to provide machine‑readable metadata and to document the presence of biometric or emotion‑analysis features so deployers can comply.

Looking ahead, the guidance is likely to accelerate procurement conversations in HR departments and push vendors to publish clearer product information. As the 2 August start date approaches, expect national market surveillance authorities to issue further operational guidance and for vendor documentation to become a focal point in employer compliance checks — a change that will reshape how HR teams sign off on chatbots, screening tools and any system that claims to read feelings at work.