Do AI-Generated Ads Need a Label in the EU? What Marketers & Designers Need to Know in 2026

Not every AI-generated or AI-assisted advertisement needs a visible AI label in the European Union.

Since 2 August 2026, Article 50 of the EU AI Act requires visible disclosure in specific situations, particularly when AI-generated or manipulated image, audio, or video content qualifies as a deepfake. AI providers also face a separate requirement around machine-readable marking of synthetic content.

For marketers and designers, that distinction matters.

Using AI somewhere in the production process does not automatically mean placing an "AI-generated" badge on every banner, product image, social ad, or resized creative.

A more useful question is:

What does the finished ad make the viewer believe is real?

That is where the EU rules become much more practical.

Important: This article explains the EU transparency rules from an advertising and creative-production perspective. It is not legal advice. Whether Article 50 applies can depend on the specific content, workflow, audience, platform, and use case.

Table of contents

• Quick answer: when do AI-generated ads need a label?
• What changed on 2 August 2026?
• The two AI transparency obligations marketers keep mixing up
• What is a deepfake under the EU AI Act?
• Which AI-created ads may need a label?
• Does AI resizing require an AI label?
• What should an EU AI label look like?
• What Google Ads is doing about AI labels
• What Meta is doing about AI-generated ads
• What TikTok is doing about AI-generated advertising
• Who is responsible: the brand, agency, designer, or AI tool?
• A practical AI-ad labelling checklist for marketers
• Frequently asked questions

Quick answer: when do AI-generated ads need a label?

For professional advertising, visible AI disclosure becomes particularly relevant when an image, video, or audio asset was generated or materially manipulated with AI and could falsely appear to show an authentic person, object, product, place, entity, or event.

That is where the EU AI Act's definition of a deepfake becomes important.

By contrast, the fact that AI helped resize an ad, re-compose its layout, clean up an image, or perform another production task does not by itself create a visible labelling requirement under Article 50(4).

There is also a separate obligation for providers of certain generative AI systems to make generated or manipulated outputs machine-readable and detectable as artificial.

These are two different transparency obligations.

What changed on 2 August 2026?

Article 50 of the EU AI Act became applicable on 2 August 2026. It introduced transparency obligations covering certain AI interactions, synthetic content, deepfakes, and AI-generated public-interest text. For advertisers, the most relevant rules are the provider-side machine-readable marking requirement and the deployer-side disclosure of deepfakes.

The EU AI Act was adopted earlier, but the transparency obligations in Article 50 began applying on 2 August 2026.

The European Commission published its final Article 50 implementation guidelines on 20 July 2026 and also established a Code of Practice on Transparency of AI-generated Content intended to provide practical ways for providers and deployers to meet the marking and labelling obligations.

The Code of Practice is voluntary.

Article 50 itself is not.

There is also a limited grace period until 2 December 2026 for the machine-readable marking obligation applying to certain generative AI systems that were already placed on the market before 2 August 2026.

That does not mean all AI labelling requirements were postponed until December.

The Commission also states that deepfake content generated before 2 August 2026 is not subject to mandatory retroactive labelling under these provisions, although voluntary disclosure is encouraged.

The two AI transparency obligations marketers keep mixing up

The EU AI Act separates machine-readable marking from visible disclosure. Providers of generative AI systems can have a technical duty to make synthetic outputs detectable, while professional users of AI can separately have to disclose certain AI-generated or manipulated content to the people who see it. One does not automatically replace the other.

The distinction can be simplified like this:

A machine-readable marker embedded in a file therefore does not automatically satisfy a deployer's visible deepfake disclosure obligation.

Likewise, putting a visible "AI" badge on an ad does not necessarily satisfy every technical obligation that may apply upstream to the provider of the AI system.

For advertising teams, it helps to think of this as two separate layers:

technical provenance and viewer-facing disclosure.

What is a deepfake under the EU AI Act?

Under the EU AI Act, a deepfake is broader than a celebrity face swap or fake political video. AI-generated or manipulated image, audio, or video content can qualify when it resembles an existing or plausibly existing person, object, place, entity, or event and could falsely appear authentic or truthful.

The European Commission's Article 50 guidance identifies three cumulative elements when assessing deepfake content.

1. Resemblance

There needs to be a sufficiently strong similarity between the synthetic content and the subject being represented.

2. The subject exists or could plausibly exist

The rule is not limited to copying one specific real person.

The Commission explains that the represented person, object, place, entity, or event can be something that exists, could plausibly exist, or could plausibly have existed in reality.

3. The content could falsely appear authentic or truthful

This is the crucial part for advertising.

The Commission says the assessment can take into account factors including:

• the level of resemblance;
• the substantive message communicated by the content;
• the intended and foreseeable context in which it will be used;
• the intended or reasonably foreseeable audience;
• and what that audience would expect to be authentic or truthful.

Context therefore matters.

An obviously fictional visual seen inside a fantasy campaign is different from a photorealistic product visual presented as though it were an accurate photograph of the real product.

This leads to a practical rule for creative teams:

The important question is not simply whether AI touched the ad. It is what the finished ad leads the viewer to believe is real.

Does every AI-generated image need an AI label?

No. The European Commission explicitly states that not all AI-generated or manipulated content needs a visible label. For images, audio, and video, the Article 50(4) deployer obligation focuses on content that qualifies as a deepfake. Other rules, including provider-side machine-readable marking, are separate.

This distinction matters because modern creative workflows use AI in many different ways.

AI can:

• extend a background;
• remove an unwanted object;
• change an aspect ratio;
• re-compose a banner;
• upscale an image;
• adjust lighting;
• create a completely synthetic person;
• fabricate a realistic event;
• alter the appearance of a product;
• replace a face;
• or generate the entire advertisement.

Treating all of these actions as legally identical would miss the way Article 50 is structured.

The final output and its context matter.

Which AI-created ads may need a label?

Whether an AI-assisted advertisement needs visible disclosure depends on the final content and how it is likely to be understood, not merely on the presence of AI somewhere in the workflow. Routine production changes raise a different transparency question from fabricating a realistic person, product, place, or event.

The following examples are practical interpretations of the Article 50 framework, not individual legal rulings.

The practical lesson is:

AI assistance is not automatically the trigger. The meaning and apparent authenticity of the finished content are much closer to the issue Article 50 is trying to address.

What about a real product in an AI-generated background?

Using AI to create or replace a background around a real product does not automatically mean the advertisement needs a visible AI label. The assessment depends on the final representation, context, audience expectations, and whether the synthetic content could falsely appear authentic or truthful.

This is one of the most relevant questions for modern advertising.

Imagine a cosmetics brand photographs its real product.

The packaging, label, dimensions, logo, colour and actual appearance remain unchanged.

AI is then used to place the product in an abstract marble environment with water, flowers and dramatic studio lighting.

Compare that with another workflow where AI:

• changes the product packaging;
• adds a feature that does not exist;
• makes the package significantly larger;
• changes the material;
• changes the amount of product;
• or otherwise makes the actual item appear materially different from reality.

Those are not the same situation.

The Commission's guidance makes context and audience expectations part of the deepfake assessment.

It also notes that AI-generated or manipulated background scenes, special effects, and standard technical production processes will not necessarily cause an audience to consider the resulting content authentic or truthful.

For advertisers, the safest approach is therefore not to create a simplistic rule such as:

"AI background = label."

Instead ask:

What is the advertisement communicating as real?

Real estate shows why the distinction matters

AI staging can move beyond aesthetic editing when it changes what a viewer believes a real property currently contains or looks like. The European Commission specifically uses an authentic photograph of an empty apartment furnished using AI as an example of partially AI-modified deepfake content in its EU icon guidance.

This is a useful example because nothing as dramatic as a fake politician is involved.

The apartment itself is real.

The photograph is real.

But AI adds realistic objects to the scene.

If the final image appears to be an authentic photograph of the furnished property, the synthetic elements become relevant to the viewer's understanding of reality.

That illustrates how broad the EU definition can be.

Deepfakes are not only about faces.

They can involve objects and places too.

Does AI resizing require an AI label?

Using AI to resize or re-compose an advertisement does not by itself mean the final ad requires visible disclosure under Article 50(4). The relevant question remains whether the resulting image, video, or audio qualifies as a deepfake, not whether automation helped reposition or adapt the creative.

This distinction matters especially in performance advertising.

One approved campaign creative might need to become:

• 1080 × 1080 for square placements;
• 1080 × 1350 for feed;
• 1080 × 1920 for Stories or Reels;
• 1200 × 628 for landscape placements;
• 300 × 250 for Google Display;
• 728 × 90 for a leaderboard;
• 160 × 600 for a skyscraper;
• and multiple additional Performance Max assets.

These transformations can require substantial composition changes without changing what the advertisement depicts.

The product might move from the right side of the creative to the centre.

The headline might wrap differently.

The CTA may need to move below the product.

Background space may need to be rebuilt to make a landscape creative work vertically.

Those are significant layout changes, and adapting one approved creative across every platform size is a production problem in its own right.

But they are fundamentally different from creating a synthetic person, fabricating an event, or materially changing how a real product appears.

For Article 50(4), evaluate the final creative and what it represents, rather than treating the mere presence of AI in the production workflow as decisive.

What is the difference between standard editing and generative manipulation?

The Commission recognises that some AI systems perform an assistive function for standard editing. Under Article 50(2), the provider-side machine-readable marking obligation does not apply in the same way when the system only performs standard editing that does not substantially alter the input or its meaning.

This provider-side exception should not be confused with the deployer's visible-disclosure obligation.

They are separate tests.

Examples of standard or assistive production work can include situations where AI improves or processes existing material without materially changing what the content communicates.

By contrast, generating a new realistic subject or substantially changing the meaning of the original content can move beyond routine editing.

For designers, this is another reason not to reduce the regulation to:

"AI used = label required."

The EU framework makes more distinctions than that.

What should an EU AI label look like?

The European Commission has created optional AI icons to help deployers disclose relevant AI-generated or manipulated content. The icons are not mandatory, and using them does not by itself prove legal compliance. When disclosure is required, it still needs to be clear, distinguishable and understandable to the person seeing the content.

The Commission currently provides three concepts:

The icons are available in different black, white and transparent variations.

Importantly, EU user testing found that recognition improved when the basic icon was accompanied by explanatory text.

That makes sense from a design perspective.

An unfamiliar symbol can be technically visible while still communicating very little.

Simple labels such as:

AI-generated

or

Modified with AI

can make the disclosure easier to understand.

The Commission also makes two important points:

The official EU icons are optional.

And:

Using an EU icon does not establish legal compliance by itself.

Where should the AI disclosure appear?

When Article 50 requires visible disclosure, it must be clear and distinguishable and available by the time a natural person is first exposed to the relevant content. The Commission's icon guidance also recommends keeping disclosures perceivable, unobstructed and available when content is reshared or downloaded where appropriate.

For designers, this turns regulation into a very practical layout problem.

A disclosure that looks perfectly readable on:

1080 × 1080

might become unusable on:

728 × 90

or:

320 × 50.

It may also be covered by:

• platform UI;
• CTA overlays;
• account information;
• dynamic cropping;
• responsive ad rendering;
• or safe-zone restrictions.

When a disclosure is required, creative teams should consider:

1. sufficient visual contrast;
2. readable size;
3. placement away from likely crop zones;
4. platform interface overlays;
5. accessibility;
6. visibility at actual served dimensions;
7. all final campaign formats, not just the master design.

Disclosure therefore becomes part of creative production, not merely a legal note added after the design is complete.

What Google Ads is doing about AI labels

Google introduced new AI labelling controls across its advertising products in July 2026. Advertisers can identify assets as AI-generated or edited, Google can automatically label certain assets generated by its own AI tools, and ads targeting jurisdictions including the EU can receive visible AI disclosures.

Google's AI label functionality applies across products including:

• Google Ads;
• Display & Video 360;
• Campaign Manager 360;
• Merchant Center;
• Google Ads Editor.

Advertisers can designate assets as AI-generated or edited.

Google says information about this use of AI can appear in its advertising-transparency interfaces.

For campaigns targeting certain jurisdictions including the European Union, visible overlays may also appear on ads.

Google also permits advertisers to add their own text or visual AI disclosure directly inside image and video creatives.

That matters for design teams because disclosure placement must survive different ad-rendering environments and aspect ratios.

Google also explicitly warns in its advertising-policy update that using its AI labelling functionality does not guarantee compliance with a particular regulation.

The platform can provide the mechanism.

The advertiser still needs to assess the applicable requirement.

What Meta is doing about AI-generated ads

Meta applies "AI info" transparency to ads created or significantly edited with generative AI and is expanding its systems to detect signals from third-party AI tools. Depending on the type of modification, disclosure information can appear through "About this ad" or more prominently alongside the advertisement.

Meta first built its advertising disclosure system around its own generative AI tools.

It has since expanded the approach.

Meta says it is beginning to detect advertisements created or edited with third-party AI systems through industry-standard signals.

Its "About this ad" experience provides a central location for additional transparency information.

For significant generative edits, Meta can apply an "AI info" disclosure.

Meta also treats photorealistic AI-generated humans more prominently in its disclosure approach.

In July 2026, Meta announced that it was signing the EU AI Act Code of Practice on Transparency of AI-Generated Content.

For advertisers, the important lesson is that the provenance of a creative may increasingly travel with the asset through technical signals rather than depending only on what an advertiser manually declares.

What TikTok is doing about AI-generated advertising

TikTok has its own AI-generated-content disclosure rules for advertising. Its platform policy can require disclaimers for images, videos, or audio that are fully AI-generated or significantly modified by AI, meaning TikTok's disclosure requirement may be broader than the minimum visible-labelling requirement under Article 50.

TikTok Ads Manager provides an AI-generated content disclaimer.

For relevant non-Spark ads, advertisers can indicate:

This ad contains AI-generated content.

TikTok's advertiser guidance covers media that is:

• completely AI-generated;
• or based on real source material that has been significantly modified with AI.

TikTok can therefore require disclosure under its own platform rules even where an advertiser might reach a different conclusion under the minimum Article 50 deepfake test.

This highlights an important distinction:

EU law and advertising-platform policy are not the same thing.

Before publishing an AI-assisted campaign, marketers should consider both:

1. the applicable legal requirement

and

2. the current policy of the destination advertising platform.

Satisfying one does not automatically satisfy the other.

Who is responsible: the brand, agency, designer, or AI tool?

Under the EU AI Act, professional users of AI systems can qualify as deployers. Where an organisation uses an AI system under its authority, the organisation can be the deployer rather than each individual employee. Responsibility therefore depends on how the AI system is professionally used and controlled, not simply on who physically clicked the generate button.

The European Commission specifically gives an advertising company as an example.

It explains that employees such as:

• digital animators;
• web designers;
• content creators;
• journalists;

working under a legal person's instructions and control should not generally be treated as separate deployers from that organisation.

The Commission also notes that a legal person can remain the deployer when contractors or freelancers operate an AI system on its behalf and under its responsibility and control.

For agencies and brands, this suggests a useful operational habit:

Do not wait until media buying begins to ask whether disclosure is needed.

Make AI provenance part of the creative workflow.

A simple campaign record might document:

• which AI systems were used;
• what was generated;
• what was modified;
• which real-world people, products, objects, places or events were affected;
• whether the final content could appear authentic;
• whether disclosure was considered necessary;
• which advertising platform will receive the creative;
• and who approved the final decision.

For ambiguous or high-risk use cases, obtain appropriate legal advice.

How AI disclosure works when exporting ads with Oppye

Oppye gives users the option to add a small visible AI disclosure badge to the banners they choose, which is then included in the exported files. The option is disabled by default because using AI somewhere in the production workflow does not automatically mean that every advertisement requires visible disclosure under Article 50.

This distinction is particularly relevant to how Oppye works.

Oppye can generate new advertising creatives with AI.

But it can also take an already approved banner and re-compose the design into multiple campaign formats while preserving important elements such as:

• product;
• logo;
• headline;
• CTA;
• colours;
• visual identity.

Those workflows raise different transparency questions.

If AI is used to generate a realistic synthetic person or materially change what a real product appears to be, visible disclosure may become relevant.

If AI is used to reorganise the same approved creative across multiple canvas sizes, AI involvement alone does not automatically turn the resulting ad into a deepfake.

That is why the AI disclosure badge in Oppye is optional and disabled by default.

When disclosure is appropriate because of:

• the content itself;
• a platform requirement;
• a client's policy;
• an organisation's internal transparency standard;
• or a legal assessment;

the user can turn the badge on from the AI label button on the Oppye whiteboard, choose which banners carry it, and adjust which corner it sits in before saving.

When enabled, the visible badge is included in the exported files.

This also moves the disclosure decision into the stage where campaign formats are actually being produced.

That matters because a label has to work across the final outputs, not only on the original master design.

What the Oppye AI badge does not mean

The badge is a creative-production and disclosure tool.

It is not a legal-compliance guarantee.

Enabling it does not automatically establish compliance with:

• Article 50;
• other parts of the EU AI Act;
• national law;
• consumer-protection law;
• or an advertising platform's policies.

Likewise, leaving it disabled does not mean that disclosure is legally unnecessary.

The person or organisation responsible for the campaign still needs to assess the final content and its intended use.

A practical AI-ad labelling checklist for marketers

Before publishing an AI-assisted advertisement in the EU, classify the final output rather than only the tool that produced it. Determine what the viewer is likely to understand as real, check whether disclosure may apply, review the destination platform's rules, and make disclosure part of production when necessary.

1. Identify how AI was used

Was AI used for:

• resizing;
• layout re-composition;
• cleanup;
• background generation;
• product generation;
• product alteration;
• synthetic models;
• face replacement;
• voice generation;
• video generation;
• or another significant modification?

2. Examine the final creative

Do not make the decision solely from the prompt or tool.

Look at what the audience will actually receive.

3. Identify what appears real

Does the creative realistically depict:

• a person;
• product;
• object;
• place;
• company or entity;
• event?

4. Ask what changed

Did AI simply help with production?

Or did it change something meaningful about what the viewer is being shown?

5. Consider the deepfake criteria

Think about:

• resemblance;
• plausibility;
• authenticity;
• context;
• message;
• audience expectations.

6. Check platform rules

Review the current requirements for:

• Google Ads;
• Meta;
• TikTok;
• LinkedIn;
• or any other destination.

Platform rules can be broader than the legal minimum.

7. Add disclosure when appropriate

Treat it as part of the design system rather than an afterthought.

8. Review every exported format

Make sure the disclosure remains:

• readable;
• visible;
• unobstructed;
• inside safe areas;
• and clear at the actual served size.

9. Record the decision

Keep a simple record of:

• what AI changed;
• whether disclosure was used;
• why;
• and who approved the decision.

10. Escalate ambiguous cases

For high-risk content involving:

• realistic synthetic people;
• endorsements;
• materially altered products;
• financial claims;
• healthcare;
• politics;
• public-interest issues;
• regulated industries;

consider obtaining specialised legal or compliance advice.

The EU AI labelling dates marketers should remember

Article 50 has applied since 2 August 2026. A limited grace period until 2 December 2026 concerns provider-side machine-readable marking for certain generative AI systems already on the market before August. It is not a general postponement of the deployer-side deepfake disclosure obligation.

Because the regulatory framework developed quickly throughout 2026, marketers should be careful when relying on articles written before the Commission's final July guidance.

Some earlier explanations were based on drafts.

What are the penalties for Article 50 violations?

The EU AI Act provides significant penalties for applicable transparency violations. The Commission's current summary states that fines can reach €15 million or 3% of a company's total worldwide annual turnover, subject to the Act's detailed penalty framework and proportionality rules.

The number sounds dramatic.

But fear of a maximum fine is not the most useful way to build a creative workflow.

The operational reality is simpler.

AI has moved from experimentation into normal advertising production.

It now appears inside:

• image generators;
• Photoshop-style editing tools;
• advertising platforms;
• video systems;
• layout tools;
• design software;
• production automation;
• resizing workflows.

Teams therefore need a repeatable process for answering:

What did AI change?

What does the viewer believe is real?

Does disclosure apply?

Does the advertising platform have additional requirements?

Will the disclosure survive every final campaign format?

That is both a compliance question and a creative-production question.

Frequently asked questions

Does every AI-generated image need an AI label in the EU?

No.

The EU AI Act does not create a blanket visible-labelling obligation for every image touched by AI.

For deployers publishing image, audio, or video content, Article 50(4) focuses on content that qualifies as a deepfake.

Provider-side machine-readable marking is a separate issue.

Do AI-generated ads need a label in the EU?

Some do.

Visible disclosure becomes particularly relevant when an AI-generated or manipulated advertisement realistically depicts a person, object, product, place, entity, or event and could falsely appear authentic or truthful.

The final content and context matter.

Does AI resizing require an AI label?

Not automatically.

Using AI to change dimensions, adapt a layout, or re-compose existing design elements does not by itself mean the result is a deepfake.

Assess the final content and the rules of the advertising platform where it will run.

Does an AI-generated background require a label?

Not automatically.

The answer depends on the final content, context, and whether the synthetic representation could falsely appear authentic or truthful to the intended audience.

Changing an aesthetic environment around a truthful product is a different issue from materially changing the product itself.

Does AI-generated real-estate staging require disclosure?

It may.

The European Commission specifically gives an authentic photograph of an empty apartment furnished using AI as an example of partially AI-modified deepfake content in its EU icon guidance.

If I generate a realistic person with AI, does it need a label?

Potentially yes.

The Commission's deepfake assessment covers content that resembles persons that exist or could plausibly exist and could falsely appear authentic.

Context and audience expectations still matter.

Are the official EU AI icons mandatory?

No.

The Commission's EU AI icons are optional.

They provide a common disclosure system, but using those exact icons is not mandatory.

Where Article 50 requires disclosure, however, the underlying disclosure requirement is not optional.

Does using the official EU icon guarantee compliance?

No.

The European Commission explicitly states that using its icons does not establish legal compliance by itself.

Can a company use its own AI badge?

Potentially, yes.

The EU icons are optional.

Whatever disclosure method is used still needs to satisfy the applicable transparency requirements.

Is invisible metadata enough for a visible deepfake disclosure?

No.

The Commission explains that deployers cannot rely only on machine-readable marking to satisfy the Article 50(4) disclosure obligation.

The disclosure needs to be understandable and perceivable by people without requiring specialised technical tools.

When did Article 50 start applying?

2 August 2026.

A limited transition until 2 December 2026 concerns machine-readable marking for certain generative AI systems already placed on the market before 2 August.

What is the maximum fine?

The EU framework provides for applicable fines reaching €15 million or 3% of worldwide annual turnover, subject to the detailed penalty rules and proportionality provisions.

Does AI-written advertising copy need an AI label?

Ordinary AI-assisted marketing copy does not automatically fall under Article 50's separate public-interest text disclosure rule.

That rule specifically concerns AI-generated or manipulated text published with the purpose of informing the public on matters of public interest.

The Act also provides an exemption where such text has undergone qualifying human review or editorial control and a person or organisation holds editorial responsibility.

The real takeaway for marketers and designers

The EU's transparency rules are narrower than "label everything made with AI," but the concept of a deepfake is broader than many marketers assume. It can involve people, objects, places, entities, and events. The best approach is therefore to assess the finished creative and make any necessary disclosure part of the production workflow.

The useful question is no longer:

"Did we use AI?"

AI is already embedded throughout creative production.

The better question is:

"What does this final advertisement lead the viewer to believe is real?"

That distinction matters.

It separates AI-assisted production from synthetic representations that can change a viewer's understanding of reality.

And when disclosure is appropriate, it needs to survive the same production process as:

• the logo;
• headline;
• product;
• CTA;
• pricing;
• and legal copy.

That is why Oppye includes an optional AI disclosure badge directly in the export workflow.

The user decides when disclosure belongs on the creative.

Oppye provides a practical way to include it when the campaign requires it.

Related reading

• Google Performance Max Banner Sizes: Image Specs, Safe Zones & Creative Checklist (2026) — useful when required information or disclosures need to remain readable across responsive Performance Max assets.
• How to Turn One Ad Into Every Size Your Campaign Needs — a practical guide to adapting one approved creative across Meta, Google Display, LinkedIn, Performance Max, and other advertising formats.
• Canva Magic Resize Alternative for Ads: How to Turn One Creative Into Every Platform Size — why campaign production often requires genuine re-composition rather than simply changing canvas dimensions.
• 10 AI Design Tools Every Graphic Designer Needs in 2026 — a practical look at where AI generation, editing, design, upscaling, and ad-production tools fit into a professional creative workflow.

Primary sources and further reading

This article prioritises current first-party regulatory and platform sources.

European Commission and EU AI Act

• Guidelines on transparency obligations for providers and deployers of AI systems
• Transparency obligations under Article 50 of the AI Act: Questions & Answers
• EU Icons for labelling AI-generated content
• Code of Practice on Transparency of AI-generated Content
• Quick Facts: Transparency rules for AI systems
• EU AI Act Article 50: Transparency obligations for providers and deployers of certain AI systems
• EU AI Act Article 99: Penalties

Advertising platforms

• Google Advertising Policies: Updates to AI labeling requirements (July 2026)
• Google Ads: Use AI content label settings and disclosures
• Meta: Expanding GenAI Transparency for Meta's Ads Products
• TikTok Ads Manager: About ad disclaimers
• TikTok Ads Manager: Add disclaimers to ads

Regulation, implementation guidance and platform policies can continue to evolve. Check the current applicable rules before relying on this article for a specific campaign or legal decision.