← All articles

redaction

AI Act: 3 days to respond — EU consultation on transparency

The European Commission closes its consultation on transparency guidelines (Article 50 AI Act) on 3 June 2026. Last call to finalize your “AI” notices and labelling of synthetic content.

On 8 May 2026, the European Commission published draft guidelines detailing how to implement the transparency obligations set out in Article 50 of Regulation (EU) 2024/1689 (AI Act). A targeted consultation is open until 3 June 2026.

The facts

The draft clarifies when and how to inform people they are interacting with AI, how to signal the use of emotion recognition or biometric categorisation tools, and how to label synthetic content (deepfakes). Overall AI Act rules start phasing in from 2 August 2026.

Legal basis

Article 50 of the AI Act covers four areas: interaction with natural persons, emotion recognition, biometric categorisation, and generation/manipulation of synthetic content. The guidelines published on 8 May 2026 aim to clarify interpretation and operational implementation ahead of entry into application.

What this means for Luxembourg businesses

  • Short timeline and direct risk: from 2 August 2026, organisations using chatbots, assistants, emotion analysis tools, or producing AI‑generated content must demonstrate clear, visible and timely transparency notices.
  • Governance required: in Luxembourg, these duties interact with GDPR (informing individuals; DPIA where needed) and, in regulated sectors, with DORA/NIS 2. Missing notices, unlabeled deepfakes, or incomplete procedures may trigger swift supervision. Build on compliant AI governance to structure your approach.
  • GDPR alignment: update your GDPR information notices and records, and prepare evidence of display/consent where applicable.

Immediate actions to take this week

  • Map your “AI touchpoints”: where users, clients, patients, applicants or staff interact with AI; any emotion analysis, biometric categorisation, or generation/manipulation of images/voice/text?
  • Draft and test contextualised notices: chatbot opening message; visible banner/overlay on synthetic content; specific wording for emotion/biometric categorisation; plan FR/EN/DE versions.
  • Plan technical labelling (watermark/metadata) and a “no deepfake without labelling” policy; define placement and persistence across channels.
  • Update notices, playbooks and controls (logging, accessibility, evidence of display).
  • Build an “Article 50” register: scope analysis, labelling/notice decisions, UI mockups, display logs and accessibility tests.
  • Contribute to the Commission’s consultation before 3 June 2026 to request sector examples and labelling format clarifications.

Need a rapid readiness check before the deadline? Reach out via our contact page.

Article generated by Luxgap regulatory watch. For tailored guidance on this topic, contact us.

LUXGAP NEWSLETTER

Get our analyses the moment they drop.

GDPR, NIS 2, AI expertise articles, plus invitations to free webinars + trainings at Luxgap. 1 to 2 emails per week max, one-click unsubscribe.

Your data is never shared. GDPR-compliant (we're DPOs after all).

A question on this topic?

Our team usually replies within one business day. Configure your quote or write to us.

Build my quote →