EU AI Act Article 50: 2026 AI Content Labeling Rules

Quick Answer

From 2 August 2026, the EU AI Act's transparency obligations (Article 50) become applicable, and they change how AI-generated advertising creative must be handled in the European Union. The duty splits by role. Providers of generative AI systems must ensure that AI-generated or manipulated content — synthetic audio, image, video and text — is marked in a machine-readable format and detectable as artificially generated or manipulated. Deployers, the businesses and agencies that actually use these systems, must disclose deepfakes, disclose AI-generated text published to inform the public on matters of public interest, and inform people when they are interacting with an AI system such as a chatbot, unless that is already obvious. In the European Commission's framing, certain AI-generated content should be clearly and visibly labelled, namely deep fakes and text published to inform the public on matters of public interest. For advertisers and agencies that produce synthetic images, voiceovers, avatars or AI-written copy, this makes disclosure a compliance requirement rather than a stylistic choice, layered on top of each platform's own AI-content rules. The transparency layer is distinct from the Act's prohibited and high-risk rules, and it sits within a broader risk-based framework. Map your exposure with the Legal Compliance Scan, audit creative libraries with the AI Compliance Audit, and track updates on the Policy Change Tracker.

Predict ad disapprovals before you submit — try the free Meta Rejection Predictor

The EU AI Act's Article 50 Transparency Rules in 2026: Labeling AI-Generated Content and Deepfakes for Advertisers

What Changes Under Article 50 in 2026

The EU AI Act is a risk-based law, and its transparency obligations — commonly cited as Article 50 — form a distinct layer of that law. According to the European Commission, the transparency rules of the AI Act come into effect in August 2026, aligning with the Act's full-applicability milestone of 2 August 2026. From that point, the way AI-generated advertising creative is produced and disclosed in the European Union is no longer purely a matter of platform policy or brand preference; it is also a statutory requirement.

The core idea is straightforward. Where AI is used to generate or manipulate the content people see, those people should be able to know that. The Commission frames this as ensuring that humans are informed when necessary to preserve trust. In practice this splits into two responsibilities: making AI-generated content technically identifiable, and clearly disclosing specific categories of it to the audience. For advertisers and agencies increasingly reliant on synthetic images, AI voiceovers, generated video and AI-written copy, the change is significant because it converts disclosure from a stylistic option into a compliance obligation.

"certain AI-generated content should be clearly and visibly labelled, namely deep fakes and text published with the purpose to inform the public on matters of public interest.
— European Commission, EU AI Act regulatory framework"

This guide explains exactly what changes under the transparency obligations, how the duties differ between providers and deployers, which content types must be labeled, how the rules fit the wider risk-based framework, and what advertisers running AI creative on platforms such as Meta, TikTok and YouTube must do to prepare. Track the transparency milestone and related changes on the Policy Change Tracker, and see how the EU's broader platform rules interact in the European Union DSA compliance guide.

Provider Duties vs Deployer Duties

The most important structural feature of the transparency obligations is that they assign different duties to different actors. The law distinguishes between providers — those who develop and place generative AI systems on the market — and deployers — the organizations that use those systems in their own operations. Most advertisers and agencies are deployers: they use a third party's generative tool to create creative. Some, especially those building or fine-tuning their own generation pipelines, may also carry provider-side considerations.

Who Must Do What

Role	Core transparency duty	Typical actor in advertising
Provider of a generative AI system	Ensure AI-generated or manipulated content — synthetic audio, image, video and text — is marked in a machine-readable format and detectable as artificially generated or manipulated	The AI vendor or model provider supplying the generation tool
Deployer of an AI system	Disclose deepfakes; disclose AI-generated text published to inform the public on matters of public interest; inform people when they interact with an AI system, such as a chatbot, unless obvious	The advertiser, brand or agency using the tool to produce creative or run a chatbot

The distinction matters because it determines who is responsible for which control. The provider-side duty is largely technical: it concerns embedding a machine-readable marker so that generated content can be detected as artificial downstream. Advertisers usually cannot satisfy that duty themselves; they inherit it through the tools they choose, which is a reason to prefer vendors that clearly support these marking obligations. The deployer-side duty is largely about audience-facing disclosure: telling people that a piece of content is a deepfake, that a public-interest text was AI-generated, or that they are talking to a machine. That is a duty the advertiser or agency must actively perform, and it cannot be outsourced to the tool. Understand the terminology behind these roles in the compliance glossary.

Which AI Content Must Be Labeled

Not every use of AI in a campaign triggers the same disclosure. The transparency obligations focus on content that could mislead people about what is real or who they are dealing with. Mapping your creative against these categories is the practical core of compliance, because it tells you where a visible label is expected and where the duty is primarily a technical marking one carried by the provider.

Content Types and Their Requirements

Content type	What the transparency layer expects	Primary responsible party
Any AI-generated or manipulated media (audio, image, video, text)	Marked in a machine-readable format and detectable as artificially generated or manipulated	Provider
Deepfakes (AI-generated or manipulated image, audio or video resembling real people, places or events)	Clearly and visibly disclosed to the audience	Deployer
AI-generated text published to inform the public on matters of public interest	Disclosed as artificially generated, in the relevant cases	Deployer
AI systems that interact with people, such as chatbots	People are informed they are interacting with an AI system, unless it is already obvious	Deployer

A few practical readings help here. First, the machine-readable marking duty is broad — it reaches AI-generated content generally — but it is a technical, provider-side obligation, so an advertiser's job is largely to use tools that support it rather than to build it. Second, the visible-disclosure duties are narrower and audience-facing, and they are the ones an advertiser or agency must actively perform: deepfakes, certain public-interest text, and chatbot interactions. Third, the "unless obvious" carve-out for chatbots is a judgment call, not a blanket exemption; where any doubt exists, disclosing is the safer course. Screen ad copy and scripts for risk terms with the Keyword Risk Checker before publishing.

How Article 50 Fits the Risk-Based Act

The transparency obligations do not stand alone. The AI Act is built on a risk-based approach that sorts AI systems into levels — from unacceptable risk (prohibited) through high-risk, then a transparency-risk tier, down to minimal or no risk. The transparency obligations sit in that framework as a distinct layer: they can apply to a system regardless of whether it is otherwise treated as high-risk, because the concern they address is specifically about people being misled about artificiality.

Where Transparency Sits

A separate layer, not a sub-rule of high-risk: transparency duties address disclosure and detectability, and they apply to relevant systems irrespective of the prohibited or high-risk classifications, so an advertiser cannot assume that because a tool is not "high-risk" the transparency rules are irrelevant.
Phased applicability: the Act's obligations have come into force in stages. As context, the rules on general-purpose AI (GPAI) models became effective in August 2025, while the transparency rules come into effect in August 2026. Advertisers should treat the August 2026 milestone as the operative date for creative disclosure.
Trust as the objective: the Commission describes these disclosure obligations as ensuring humans are informed when necessary to preserve trust — a framing that helps interpret borderline cases in favor of disclosure.

As general statutory context, the AI Act provides for significant penalties for non-compliance, with fines that can reach a percentage of a company's total worldwide annual turnover; the precise figures depend on the nature of the breach and should be confirmed against official EU sources rather than assumed. The takeaway for advertisers is that transparency is not a soft, best-effort expectation bolted onto the interesting parts of the law — it is a substantive obligation with enforcement behind it. Audit how your AI-assisted assets and systems map to these tiers with the AI Compliance Audit.

AI Creative on Meta, TikTok and YouTube

For most advertisers, AI-generated creative does not reach audiences directly; it reaches them through platforms. Meta, TikTok and YouTube all operate at scale in the European Union, and each has moved toward requiring or enabling disclosure of AI-generated and manipulated content. That creates a two-layer reality for advertisers: the EU AI Act's transparency obligations set the legal baseline, and each platform's own AI-content rules sit on top, sometimes with their own labels, self-declaration prompts and detection systems.

The Two-Layer Compliance Reality

Legal baseline: the transparency obligations define, as a matter of EU law, when AI-generated content must be identifiable and when deepfakes, certain public-interest text and chatbot interactions must be disclosed.
Platform layer: advertising platforms independently ask advertisers to declare AI-generated or altered media and may apply their own labels; complying with a platform's disclosure prompt is not automatically the same as satisfying the legal duty, and vice versa, so both must be checked.
Provider dependence: because the machine-readable marking duty is provider-side, the detectability of your creative depends partly on the generation tools you use; tools that embed machine-readable markers make downstream platform detection and your own disclosure easier to defend.

Practically, an agency shipping an AI-generated video ad into the EU should assume it must be both technically identifiable as AI-generated (a duty it largely inherits from its tools) and, where it is a deepfake or otherwise in scope, clearly disclosed to viewers (a duty it must perform). Where a platform offers an AI-content self-declaration, use it, but treat it as one control among several rather than a complete answer to the legal obligation. Confirm the specifics of each platform's current AI-content rules directly, since platform policies evolve independently of the statute, and map cross-jurisdiction exposure with the Legal Compliance Scan.

What Advertisers and Agencies Must Do

Compliance with the transparency obligations is concrete and achievable, but it requires deliberate process rather than ad-hoc judgment on each asset. The goal is to ensure that AI-generated creative reaching EU audiences is identifiable where required and clearly disclosed where required, and that you can show your work if questions arise.

The Compliance Actions

Inventory AI use across creative: identify where generative AI produces or manipulates images, audio, video, avatars or copy in your campaigns, including partial edits and voice cloning, not just fully synthetic assets.
Choose tools that support marking: because machine-readable marking is a provider-side duty, prefer generation vendors that embed machine-readable markers and can attest to how they meet the obligation, so detectability travels with your content.
Disclose deepfakes and in-scope text: where creative is a deepfake, or is AI-generated text published to inform the public on matters of public interest, apply clear, visible disclosure to the audience.
Flag chatbots and AI agents: where you deploy chatbots or conversational AI, inform users they are interacting with a machine unless it is already obvious.
Reconcile with platform rules: satisfy each platform's AI-content declaration requirements in addition to the legal duty, and do not treat one as a substitute for the other.
Document the process: keep records of which assets used AI, which tools, what marking they support and how disclosure was applied, as evidence of good-faith compliance.

Because the obligations become applicable at the August 2026 milestone, the sensible posture is to have these controls in place before that date rather than scrambling afterward — production pipelines, vendor contracts and disclosure templates all take time to change. Treat platform policy and the legal obligation as distinct but overlapping, verify specifics against official EU and platform sources, and remember that this guide is not legal advice; where stakes are high, confirm your interpretation with qualified counsel. Standardize recurring checks with the AI Compliance Audit and monitor the milestone on the Policy Change Tracker.

AI-Generated Creative Compliance Checklist

[ ] Inventoried every campaign asset that uses generative or manipulative AI, including partial edits
[ ] Confirmed your generation tools support machine-readable marking of AI-generated content
[ ] Identified which assets are deepfakes requiring clear, visible disclosure
[ ] Identified any AI-generated text published to inform the public on matters of public interest
[ ] Added clear disclosure that users are interacting with an AI system for any chatbot or conversational agent, unless obvious
[ ] Completed each platform's AI-content self-declaration for Meta, TikTok, YouTube and others in use
[ ] Reconciled platform declarations with the legal disclosure duty rather than treating either as sufficient alone
[ ] Documented AI usage, tools, marking support and disclosures applied per asset
[ ] Set the August 2026 applicability milestone as the deadline for controls to be live
[ ] Verified interpretation against official EU sources and, where stakes are high, qualified legal counsel

Frequently Asked Questions

When do the EU AI Act's Article 50 transparency obligations become applicable, and what exactly changes?

The EU AI Act's transparency obligations, commonly cited as Article 50, become applicable at the Act's full-applicability milestone of 2 August 2026 — the European Commission states that the transparency rules of the AI Act come into effect in August 2026. From that point, the handling of AI-generated advertising creative in the European Union is governed not only by platform policy and brand preference but also by statutory duty, and that is the central change. What changes, concretely, is that two things become legally required rather than optional. First, AI-generated or manipulated content must be made identifiable: providers of generative AI systems must ensure that synthetic audio, image, video and text is marked in a machine-readable format and detectable as artificially generated or manipulated. Second, specific categories of content must be clearly disclosed to the audience: deepfakes must be disclosed, AI-generated text published to inform the public on matters of public interest must be disclosed in the relevant cases, and people must be informed when they are interacting with an AI system such as a chatbot, unless that is already obvious. In the Commission's own framing, certain AI-generated content should be clearly and visibly labelled, namely deep fakes and text published to inform the public on matters of public interest, and the broader objective is to ensure that humans are informed when necessary to preserve trust. It is worth being precise about what does not change. The transparency obligations are a distinct layer of a risk-based law; they are not the same as the Act's prohibited-use or high-risk rules, and they can apply to a system independently of whether it is otherwise classified as high-risk. As phasing context, the Act's rules on general-purpose AI (GPAI) models became effective in August 2025, ahead of the transparency milestone in August 2026, so advertisers should treat the 2026 date as the operative one for creative disclosure. For advertisers and agencies, the practical meaning is that disclosure of AI-generated creative shifts from a stylistic choice to a compliance requirement, layered on top of each platform's own AI-content rules, and the controls to satisfy it — tool selection, disclosure templates, documentation — should be in place before the milestone rather than after. Track the milestone and related updates on the Policy Change Tracker, and see how the EU's platform rules interact in the European Union DSA compliance guide. The organizing principle is that from August 2026 the transparency obligations make AI-generated content identifiability and specific disclosures a legal duty in the EU.

What is the difference between what providers and deployers must do under the transparency obligations?

The transparency obligations assign different duties to providers and to deployers, and understanding which role you occupy is the key to knowing which controls you are responsible for. A provider, in the Act's structure, is an actor that develops and places a generative AI system on the market; a deployer is an organization that uses such a system in its own operations. Most advertisers and agencies are deployers — they use a third-party generative tool to create creative — while the AI vendor supplying the tool is typically the provider. The provider-side duty is primarily technical. Providers of generative AI systems must ensure that AI-generated or manipulated content — synthetic audio, image, video and text — is marked in a machine-readable format and detectable as artificially generated or manipulated. In practice this means embedding a machine-readable marker so that generated content can be recognized as artificial downstream, by platforms and other systems. An advertiser usually cannot satisfy this duty independently; it inherits the benefit of it through the tools it chooses. That is precisely why tool selection matters so much: preferring vendors that clearly support machine-readable marking means detectability travels with your content, which strengthens both platform-level detection and your own ability to demonstrate compliance. The deployer-side duty is primarily about audience-facing disclosure. Deployers must disclose deepfakes; disclose AI-generated text published to inform the public on matters of public interest, in the relevant cases; and inform people when they are interacting with an AI system, such as a chatbot, unless that is already obvious. These are duties the advertiser or agency must actively perform on the content and experiences it publishes — they cannot be outsourced to the tool, because they concern what the audience is told. The practical consequence of the split is a division of labor: the provider is responsible for making content technically markable and detectable, and the deployer is responsible for the visible, human-facing disclosures. An advertiser that assumes its vendor's marking also handles audience disclosure has misread the structure, and so has one that applies visible labels but uses a tool that does not support machine-readable marking. Both layers matter, and they are held by different parties. Some sophisticated advertisers that build or fine-tune their own generation pipelines may take on provider-side considerations as well, which raises the stakes of getting the technical marking right. To keep these roles straight, review the definitions in the compliance glossary and audit which duties attach to your setup with the AI Compliance Audit. The organizing principle is that providers make AI content markable and detectable, while deployers perform the visible disclosures — and advertisers are usually deployers who inherit marking from their tools.

Which types of AI-generated content must be labeled or disclosed?

The transparency obligations focus on content that could mislead people about what is real or who they are dealing with, and mapping your creative against these categories is the practical core of compliance. There are, broadly, four relevant categories, and they differ in whether the duty is a technical marking obligation carried by the provider or a visible disclosure obligation performed by the deployer. The first category is AI-generated or manipulated media in general — synthetic audio, image, video and text. For this, the expectation is that content is marked in a machine-readable format and detectable as artificially generated or manipulated. This is broad in scope but technical in nature, and it is a provider-side duty; the advertiser's role is largely to use tools that support the marking rather than to build it. The second category is deepfakes — AI-generated or manipulated image, audio or video that resembles real people, places or events. Deepfakes must be clearly and visibly disclosed to the audience, and this is a deployer duty the advertiser must actively perform. The European Commission's framing is explicit that certain AI-generated content should be clearly and visibly labelled, namely deep fakes and text published to inform the public on matters of public interest. The third category follows directly from that: AI-generated text published to inform the public on matters of public interest must, in the relevant cases, be disclosed as artificially generated. This is narrower than all AI-written copy — it targets text with a public-information purpose — but advertisers producing content that informs the public on matters of public interest should treat it as in scope. The fourth category is interactive AI: systems such as chatbots that engage with people. Here, people must be informed that they are interacting with an AI system, unless that is already obvious. The "unless obvious" carve-out is a judgment call rather than a blanket exemption, and where any doubt exists, disclosing is the safer course. A useful way to operationalize this is to separate the technical from the visible. The machine-readable marking duty is broad but handled largely through tool choice; the visible-disclosure duties — deepfakes, in-scope public-interest text, and chatbot interactions — are the ones an advertiser or agency must consciously apply to what it publishes. Before publishing, screen ad copy and scripts for sensitive or misleading claims with the Keyword Risk Checker, and confirm ambiguous cases against official EU guidance rather than guessing. The organizing principle is that general AI media must be machine-readably marked (provider duty), while deepfakes, certain public-interest text and chatbot interactions must be visibly disclosed (deployer duty).

How do the transparency rules affect advertisers running AI-generated creative on Meta, TikTok and YouTube?

For most advertisers, AI-generated creative reaches audiences through platforms rather than directly, so the transparency obligations create a two-layer reality on services like Meta, TikTok and YouTube: the EU AI Act sets the legal baseline, and each platform's own AI-content rules sit on top. All three platforms operate at scale in the European Union and have moved toward requiring or enabling disclosure of AI-generated and manipulated content, which means advertisers must satisfy both layers rather than assuming one covers the other. The legal layer defines, as a matter of EU law, when AI-generated content must be identifiable and when deepfakes, certain public-interest text and chatbot interactions must be disclosed. This baseline does not depend on any individual platform's product decisions; it applies to the content itself. The platform layer is separate and platform-specific: advertising platforms independently ask advertisers to declare AI-generated or altered media, may apply their own visible labels, and may run their own detection systems. Crucially, completing a platform's AI-content self-declaration is not automatically the same as satisfying the legal duty, and satisfying the legal duty does not automatically discharge a platform's declaration requirement. Both must be checked. A third factor ties the two layers together: provider dependence. Because the machine-readable marking duty is provider-side, the detectability of your creative depends partly on the generation tools you use. Tools that embed machine-readable markers make downstream platform detection more reliable and make your own disclosure easier to defend, whereas tools that do not may leave your content harder to identify as AI-generated — a weakness on both the legal and platform fronts. Practically, an agency shipping an AI-generated video ad into the EU should assume the asset must be both technically identifiable as AI-generated, a duty it largely inherits from its tools, and, where it is a deepfake or otherwise in scope, clearly disclosed to viewers, a duty it must perform. Where a platform offers an AI-content self-declaration, use it — but treat it as one control among several rather than a complete answer to the statutory obligation. Because platform policies evolve independently of the statute and of one another, confirm each platform's current AI-content rules directly rather than relying on a general impression, and re-check periodically. Map your exposure across platforms and jurisdictions with the Legal Compliance Scan, and understand how these rules sit alongside the EU's platform obligations in the European Union DSA compliance guide. The organizing principle is that advertisers must satisfy both the AI Act's legal disclosure baseline and each platform's own AI-content rules, with tool choice affecting how well their creative meets both.

What are the risks of non-compliance, and how should advertisers prepare before August 2026?

The risks of non-compliance with the transparency obligations operate on two levels — statutory exposure under the AI Act and operational exposure on the platforms — and the sensible way to manage both is to have controls in place before the August 2026 applicability milestone rather than after. On the statutory side, the AI Act is a binding EU law, not a voluntary code, and as general context it provides for significant penalties for non-compliance, with fines that can reach a percentage of a company's total worldwide annual turnover. The precise figures depend on the nature of the breach and should be confirmed against official EU sources rather than assumed, but the direction is clear: transparency is a substantive obligation with enforcement behind it, not a soft expectation. On the operational side, advertisers also face platform-level risk, because Meta, TikTok, YouTube and other services enforce their own AI-content rules; failing to declare or disclose AI-generated media can lead to platform actions on the content or account independently of any regulatory consequence. The two risks compound: a single non-disclosed deepfake ad could be both a platform-policy problem and a legal-compliance problem at once. Preparation should be process-driven rather than asset-by-asset improvisation. Start by inventorying where generative AI produces or manipulates images, audio, video, avatars or copy across your campaigns, including partial edits and voice cloning, not just fully synthetic assets, because partial manipulation can still fall within scope. Next, choose generation tools that support machine-readable marking and can attest to how they meet the obligation, so detectability travels with your content; this addresses the provider-side duty you inherit. Then apply the deployer-side disclosures you must perform: clear, visible disclosure for deepfakes; disclosure for AI-generated text published to inform the public on matters of public interest; and clear notice that users are interacting with an AI system for any chatbot, unless obvious. Reconcile all of this with each platform's own AI-content declaration requirements, treating neither the legal duty nor the platform declaration as a substitute for the other. Finally, document everything — which assets used AI, which tools, what marking they support, and how disclosure was applied — because that record is your evidence of good-faith compliance if questions arise. Because production pipelines, vendor contracts and disclosure templates all take time to change, treat August 2026 as the date by which these controls must be live, and verify your interpretation against official EU and platform sources; where stakes are high, confirm with qualified legal counsel, as this guide is not legal advice. Standardize recurring checks with the AI Compliance Audit and monitor the milestone on the Policy Change Tracker. The organizing principle is that non-compliance carries both statutory and platform risk, so advertisers should build and document transparency controls before the August 2026 milestone.

Don't miss the next policy change.

Create a free account — track every policy change across 8 platforms, get instant alerts, and access every free compliance tool. Or try our Meta Rejection Predictor first.

Create Free Account

Report Keywords — Run AI Compliance Audit

#EU AI Act#Article 50#AI Transparency#AI-Generated Content#Deepfakes#Content Disclosure#Synthetic Media#EU Regulation#Advertisers#Agencies#2026 Policy#Compliance Guide 2026

Share This Report

Tweet Share

New York's Synthetic Performer Disclosure Law in 2026: What Advertisers Must Disclose About AI-Generated Performers

New York's first-in-the-nation law requires advertisers to conspicuously disclose AI-generated synthetic performers in ads — a direct advertiser duty, effective June 2026.

South Korea's AI Advertising Law in 2026: Labeling Rules, KFTC Enforcement and Misleading-Claim Penalties

From 2026 South Korea moves to require labeling of AI-generated advertising and tightens rules on misleading claims and virtual endorsers. Here is what changes and how advertisers comply.

COPPA Rule Amendments in 2026: The Children's Privacy Overhaul That Changes Targeted Advertising, Biometrics and Data Retention

The FTC's amended COPPA Rule reaches full compliance in April 2026 — requiring separate parental consent for third-party targeted advertising and limiting data retention.

Platforms

Industries

Knowledge

Blog

The EU AI Act's Article 50 Transparency Rules in 2026: Labeling AI-Generated Content and Deepfakes for Advertisers