EU Publishes AI Content Labelling Playbook Ahead of August Deadline

By Hector Herrera | June 16, 2026 | Government

The European Commission has released a voluntary Code of Practice for AI content labelling, giving companies a compliance roadmap ahead of August 2, 2026 — the date mandatory AI transparency requirements kick in under the EU AI Act. Starting that day, any company serving EU users must visibly label deepfakes, AI-generated text on public-interest topics, and interactive AI systems that could be mistaken for humans.

The voluntary code matters because the underlying law does not. Article 50 of the EU AI Act imposes labelling obligations whether a company signs the code or not. The code simply tells companies how to meet those obligations in a way the Commission considers acceptable.

What the Rules Actually Require

Under Article 50, three categories of AI output must be labelled starting August 2:

• Deepfakes — synthetic video or audio depicting real people must be clearly identified as artificially generated or manipulated.
• AI-generated text on public-interest topics — news, politics, science, and similar domains require disclosure when text is machine-written.
• AI chatbots and virtual assistants — systems that interact with users must disclose that the user is not talking to a human.

The Commission's Code splits implementation between two layers of the AI supply chain. Model developers are responsible for embedding machine-readable metadata into outputs — a technical signal that downstream systems can detect. Deployers — the companies building products on top of those models — must surface visible labels to end users.

Why This Is Harder Than It Sounds

Machine-readable tagging requires that content provenance travel with the file or text as it moves across platforms. That's technically solvable through standards like C2PA (Coalition for Content Provenance and Authenticity), which embeds cryptographic metadata into images and video. But text provenance — particularly when AI-generated content is copy-pasted, reformatted, or translated — breaks those chains quickly.

Deployers face a different challenge: what counts as "visible" disclosure? The Code offers guidance, but enforcement will fall to national AI authorities, and interpretation may vary across the EU's 27 member states during the early months.

What Businesses Need to Do Now

Companies with EU users have roughly six weeks. The immediate checklist:

1. Audit your AI outputs — identify which of your AI-generated content falls into the three mandatory categories.
2. Talk to your model providers — confirm they are building machine-readable metadata into outputs before August 2.
3. Design user-facing labels — plan how disclosures will appear in your UI without burying them in terms-of-service fine print.
4. Don't wait for sign-off on the voluntary code — the mandatory obligations under Article 50 apply regardless.

Penalties under the EU AI Act for transparency violations reach €15 million or 3% of global annual turnover, whichever is higher.

What to Watch

The Commission has not yet named the national authorities that will handle enforcement in each member state, and several countries are still setting up their oversight structures. Watch for enforcement guidance from larger markets — Germany, France, and the Netherlands — in July, which will signal how aggressively Article 50 will be applied from day one.