AI Act: July 22 — last chance to sign the AI transparency Code

Critical deadline: the European Commission will publish an initial list of signatories to the Code of Practice on Transparency of AI‑Generated Content before August 2, 2026. To be included, submit the signature form by July 22, 2026 at 18:00 (CEST).

The facts

On July 9, 2026, the Commission and the AI Board confirmed that the Code of Practice on transparency of AI‑generated content is an “adequate” voluntary tool to meet the AI Act’s transparency duties on user information, labelling and technical marking. As of August 2, 2026, these obligations apply across the EU under the AI Office’s central enforcement.

Legal basis

• AI Act — Article 50: informing people when they interact with AI; labelling synthetic content (including deepfakes); and technical marking/detection where feasible. For a recap, see the AI Act obligations.
• Article 50(7): mandate to the AI Office to facilitate EU codes of practice detailing identification, marking and labelling modalities.
• Governance: role of the AI Board and AI Office in interpretation, coordination and publication of signatories.
• Procedural deadline: signature forms due by July 22, 2026, 18:00 (CEST) to appear on the initial list.

What this means for Luxembourg companies

• Who is in scope: any entity in Luxembourg deploying generative systems (text, image, audio, video) accessible in the EU, or publishing synthetic content externally or internally.
• Risk and timing: on August 2, 2026, informing users and labelling AI‑generated/manipulated content becomes enforceable. Not being ready invites scrutiny. Signing the Code and showing concrete measures in the week of July 22 signals compliance.

Practical actions to take this week

1) Map AI touchpoints and decide to sign

List every use case where users see or interact with AI content (chatbots, automated emails, offers, images/videos, audio summaries, avatars). Prepare governance (executive sponsor, DPO/CISO, legal) and file the form by July 22, 2026, 18:00 (CEST).

2) Implement notices and markers aligned with Article 50

• Visible transparency: a notice at the user touchpoint (e.g., “AI‑generated content”) with a “Learn more” link.
• Technical marking: watermarking, C2PA/Content Credentials, persistent metadata; where feasible, machine‑detectable signals (audio/video). Document technical limits where marking is unreliable.

3) Define a labelling standard and publishing control

Standardize a labelling schema (short text, pictogram, link) and a pre‑publication playbook (who checks, with what tool, what timestamped evidence). Retain evidence (screenshots, hashes, metadata) for audit.

4) Update internal policies and employee information

Embed AI transparency duties into editorial charters, marketing policies and social media guides; update notices on internal HR/IT chatbots.

5) Prepare the public “AI Transparency” page

Describe when content is labelled, how it is marked, known technical limits, and how to report an issue. Link your signatory status once public.

6) Anticipate requests and inspections

Train support and comms to answer about the label, detection and training data. Prepare an evidence dossier for the AI Office and national authorities.

7) Finalize the signature file

Validate governance, sign at executive level, and submit by July 22, 2026, 18:00 (CEST). To accelerate operational alignment, leverage our expertise in AI governance and compliance.

Accelerate in Luxembourg

If you need to align processes with the AI Act within days, mobilize DPO, legal and security teams and prioritize high‑exposure public flows. Our page on AI compliance in Luxembourg highlights local levers.

If you need immediate support for labelling, technical marking and your “AI Transparency” page, you can get in touch with Luxgap.

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