Labeling AI-generated content is about to become a legal requirement in the European Union. In June 2026, the European Commission published a voluntary Code of Practice to help organizations meet the AI Act transparency obligations that apply from 2 August 2026. This article explains what must be labeled, what the new Code does, and what it means for any organization whose AI content reaches the EU.
What the EU AI Act Requires for Labeling AI-Generated Content
The requirement comes from Article 50 of the AI Act, the section that sets transparency obligations for providers and deployers of generative and interactive AI systems. These obligations arrive on two dates, following the digital omnibus approved by the European Parliament in June 2026.
From 2 August 2026, deployers must clearly label deepfakes and AI-generated or AI-manipulated text published to inform the public on matters of public interest, and anyone deploying an interactive system such as a chatbot must tell people they are interacting with AI rather than a human. The provider duty to mark AI-generated or AI-manipulated audio, image, video, and text in a machine-readable format, so the content can be detected as artificial, applies from 2 December 2026 for generative systems already on the market before 2 August 2026, with systems launched after that date expected to comply right away.
These rules sit within the broader risk-based structure of the AI Act. For how that structure compares with other frameworks, see the guide on EU AI Act compliance alongside NIST and ISO 42001.
What the New Code of Practice Does, and Does Not Do
On 10 June 2026, the European Commission published the final Code of Practice on marking and labeling of AI-generated content. It was drafted by independent experts through a stakeholder process and is organized into two parts: one for providers, covering machine-readable marking and detection, and one for deployers, covering the labeling of deepfakes and public-interest text. Taken together, it is the EU’s most concrete guidance yet on labeling AI-generated content.
The critical point is what the Code is not. It is voluntary, and signing or following it does not by itself prove compliance with the AI Act. It is a practical route to meeting the obligations, not a substitute for them. Organizations remain responsible for satisfying Article 50 whether or not they adopt the Code.
What Content Must Be Labeled
The obligations target the content most likely to deceive:
- Deepfakes. AI-generated or AI-manipulated image, audio, and video that resembles real people, places, or events must be clearly disclosed as artificial.
- Public-interest text. AI-generated or AI-manipulated text published to inform the public on matters of public interest must be labeled.
- Synthetic media at the source. Providers must embed machine-readable markings so AI output can be detected downstream.
- Interactive AI. People must be told when they are dealing with a chatbot or similar system rather than a person.
There are exceptions. Labeling obligations are eased where AI-generated text has undergone human review and sits under editorial responsibility, and where AI performs an assistive or minor editing function. The exact boundaries are part of what the Commission’s accompanying guidelines aim to clarify.
The EU Icons for AI-Generated Content
To make disclosure consistent, the Commission has published a set of standardized icons for marking AI-generated content. Standard marks matter because labeling only works if readers, regulators, and courts recognize and trust the same signal. Organizations can review the official set on the Commission’s EU Icons for labelling AI-generated content page.
Does This Apply to US Companies?
Often, yes. The AI Act reaches beyond the borders of the EU. A company based in the United States can fall within scope when it places a generative AI system on the EU market, or when the AI-generated output it produces is used in the EU, regardless of where the company itself is located. For many US organizations publishing content, running chatbots, or shipping AI features that reach European users, the transparency obligations are not optional and the 2 August 2026 date is real.
Because applicability turns on specific facts, organizations should assess their own exposure rather than assume distance from Europe puts them outside the rules.
How to Prepare for Labeling AI-Generated Content
Preparing for labeling AI-generated content is less about a single tool and more about knowing where AI touches the content you publish:
- Inventory where you generate and publish AI content. Map every system that creates or alters audio, image, video, or text.
- Identify what triggers a label. Flag deepfakes, public-interest text, and chatbots across your operations.
- Implement marking and visible labels. Apply machine-readable marking at the source and clear, consistent disclosures to users.
- Preserve provenance. Keep records of what was generated, how, and when, before anyone asks for them.
- Fold it into your AI governance program. Treat transparency as one control within a managed system, not a one-off task. The legal dimensions are covered in this guide to ISO 42001 and AI legal risk, and the wider set of standards lives in the Elevate Framework Library.
Elevate Consult helps organizations turn AI transparency rules into a working program. The ISO 42001 AI Governance Readiness Bundle is a structured place to start.
How Elevate Consult Helps
Elevate Consult helps providers and deployers prepare for the AI Act transparency obligations as part of a governance program aligned to ISO 42001 and the NIST AI Risk Management Framework. The work spans AI inventory, policy, labeling and provenance controls, and the evidence that demonstrates a good-faith path to compliance.
Organizations that want to be ready before 2 August 2026 can start a conversation with the Elevate team.
Key Takeaways
- Labeling AI-generated content becomes a legal requirement in the EU on 2 August 2026 under Article 50 of the AI Act.
- Providers must mark AI output in a machine-readable way; deployers must label deepfakes and public-interest text and disclose chatbots.
- The European Commission’s Code of Practice, published 10 June 2026, is voluntary and supports compliance but does not establish it.
- The rules can reach US companies whose AI systems or AI-generated output are used in the EU.
- Standardized EU icons exist to make disclosure consistent and recognizable.
Frequently Asked Questions
What does the EU AI Act require for labeling AI-generated content?
Under Article 50 of the AI Act, deployers must clearly label deepfakes and AI-generated text published on matters of public interest, and people must be told when they are interacting with an AI system such as a chatbot. These deployer and disclosure duties apply from 2 August 2026. The separate provider duty to embed machine-readable marking in AI-generated audio, image, video, and text applies from 2 December 2026 under the June 2026 digital omnibus, for generative systems already on the market before 2 August 2026.
When do the EU AI Act transparency rules apply?
Most apply from 2 August 2026, including the duty to label deepfakes and AI-generated public-interest text and to disclose AI chatbots. The provider duty to embed machine-readable marking in AI-generated content was moved to 2 December 2026 by the digital omnibus approved in June 2026, with a grandfathering allowance for generative systems already on the market before 2 August 2026.
Is the EU Code of Practice on AI labeling mandatory?
No. The Code of Practice published on 10 June 2026 is voluntary. Following or signing it can support a demonstration of compliance with the AI Act, but it does not by itself establish compliance. The underlying Article 50 obligations are mandatory regardless of whether an organization adopts the Code.
Does the EU AI Act apply to US companies?
It can. The AI Act applies beyond the EU when a company places a generative AI system on the EU market or when the AI-generated output it produces is used in the EU, regardless of where the company is based. US organizations whose AI content or systems reach European users should assess whether the transparency obligations apply to them.
What content must be labeled as AI-generated?
The obligations cover deepfakes, meaning AI-generated or AI-manipulated image, audio, and video resembling real people or events, as well as AI-generated text published to inform the public on matters of public interest. Providers must also embed machine-readable marking in synthetic media, and deployers must disclose interactive AI such as chatbots. Limited exceptions apply where content is under human editorial review.