EU AI Act Enforcement Begins: Article 50 Labels

EU AI Act enforcement and Article 50 transparency

Article 50 of the AI Act sets transparency duties for providers and deployers of generative AI systems. The duties cover marking and detecting AI-generated content and labeling deepfakes.

A provider builds or supplies the AI system. A deployer uses it, for example by publishing its output.

Since 2 August 2026 the Commission's AI Office, together with national authorities, has been enforcing the Act, and the Article 50 rules have applied since the same day. For a business that runs a chatbot or publishes AI-made images, video, audio or text, the duties are now enforceable.

AI Act deepfake labeling and watermarking rules

The Commission names three rules. The table shows who each one covers.

The three Article 50 labeling rules that apply from 2 August 2026

Rule	Who it covers	What it requires
Say it is AI	Chatbots and other interactive AI systems	Tell users they are dealing with AI, not a human
Label deepfakes	Images, videos or audio edited or generated using AI	Label the content as a deepfake
Add machine-readable marks	AI-generated or altered content	Carry marks so the content can be detected more easily

The Commission says the measures are meant to reduce deception and manipulation and help people make informed choices. It adds that they give businesses clearer obligations and a practical way to show compliance. The schedule for the other AI Act duties, including the high-risk delay, is in EU AI Act 2026: What Applies in August, What Waits.

AI Act 2 August 2026 and the code of practice signatory list

On 2 August 2026 the Commission published a first list of more than 180 organizations that signed the transparency code. The code's own policy page puts the count at about 190 organizations by the end of July 2026. Both numbers are reported here as each page gives them: "more than 180" is a lower bound, and neither page says how the two counts relate.

EU AI Act code of practice for AI content labeling

The code of practice is a voluntary tool for meeting the rules in the table above. Independent experts drew it up in a multi-stakeholder process coordinated by the AI Office. It helps providers and deployers of generative AI systems comply with the labeling and marking duties in Article 50(2), (4) and (5) of the AI Act.

Article 50 code of practice: providers and deployers

The code has two sections, one for each role. Each section covers one side of the Article 50 duties:

• Section 1 is for providers. It sets rules for marking and detecting AI-generated and manipulated content.
• Section 2 is for deployers. It sets rules for labeling deepfakes and AI-generated and manipulated text.

In short, providers mark the content and deployers label it. The EU has created a set of icons that deployers of generative AI systems may use to label their AI-generated content. Signatories will also work together in Signatory Taskforces, yet to be set up, to share practices and advance marking and labeling.

Signing the code of practice and Article 50 compliance

The Commission and the AI Board have confirmed that the code is an adequate voluntary tool to demonstrate compliance with the transparency obligations. Following a positive assessment, providers and deployers who sign can rely on the code's measures to demonstrate compliance. The Commission says this reduces administrative burden and gives predictability, legal certainty and trust across all Member States.

Signing does not remove the duty. Adherence to the code is voluntary, while the Article 50 requirements are legal obligations. A signature changes how you show compliance, not whether you owe it.

Without signing: AI Act transparency requirements

A provider or deployer can comply through other means. If it does, it has to demonstrate that those measures are adequate. Different market surveillance authorities assess this individually.

The code is published as a PDF. To work with specific provisions, you can convert it to Markdown first.

AI Act enforcement: where to report Article 50 transparency breaches

Alongside the start of enforcement, the Commission links three channels for reporting a breach of the Act. Each one handles a different kind of report.

AI Act transparency obligations: three reporting channels

Reporting channels the Commission links with AI Act enforcement

Channel	Meant for
AI Act complaints tool	Complaints about AI Act breaches
AI Act Whistleblower Tool	Whistleblower reports
Complaints channel for downstream providers	Downstream providers using general-purpose AI models

Two of the channels take complaints and whistleblower reports; the third is for downstream providers that use general-purpose AI models. The same page also points to further guidance, including the Commission's guidelines on AI-generated content transparency and on general-purpose AI model providers.

Outside the EU: check Article 50 and AI content labeling

A business outside the EU that serves EU users has two checks to make. Both come from the provider and deployer roles in Article 50:

• Has the generative AI provider you build on signed the code? If so, it can rely on the code's marking measures to demonstrate its own compliance.
• Do you publish AI-generated content? Then you act as a deployer, and Section 2 of the code is the part written for deployers. Deployers of generative AI systems may use the EU icons to label that content.

For how the schedule moved, including the delay to high-risk obligations, see EU AI Act Omnibus: New Deadlines and Nudifier Ban. EU AI Act 2026: What Applies in August, What Waits sets out what applies in August and what waits.

Summary: code of practice signatures and AI content labeling

EU AI Act enforcement began on 2 August 2026, and the Article 50 labeling rules apply from the same day. Signing the code of practice is voluntary and changes how you show compliance, not the legal duty. Check whether your AI provider signed the code, and check your own labeling against Section 2 if you publish AI-generated content.