Introduction
Since August 2, 2026, the use of artificial intelligence in advertising must be assessed at two levels:
- The first concerns transparency obligation relating to the use of AI, as provided for in Article 50 of the EU Artificial Intelligence Act, certain provisions of which entered into application on that date ;
- The second covers the rules that already apply to commercial communications, including the accuracy of claims, intellectual property, image rights, data protection and contractual responsibility.
However, not every use of AI needs to be disclosed to the public. The transparency obligation primarily applies to content constituting deepfakes, as well as to certain texts concerning matters of public interest. Standard retouching, background removal or preparatory assistance therefore do not automatically trigger a disclosure requirement. Conversely, the absence of such a disclosure never exempts the advertiser from complying with the rules governing misleading commercial practices or with third-party rights.
When must advertising using AI be disclosed?
Article 50 distinguishes providers from deployers
The EU Artificial Intelligence Act, Regulation (EU) 2024/1689 allocates duties between the provider of an AI system and its deployer.
- A provider develops the system or markets it under its own name. It must enable the identification of synthetic output through machine-readable marking, except where the system merely performs standard editing or does not substantially alter the input data or its meaning.
- A deployer uses the system under its authority for professional purposes. A trademark, advertising agency or production company may therefore be a deployer. Employees acting under a company’s control are not separate deployers.
Contractual allocation remains important, but it cannot override the legal qualification resulting from the parties’ actual control over the advertising process.
Advertising deepfakes require visible disclosure
Under the European Commission Guidelines published on July 20, 2026, deployers must disclose AI-generated or manipulated image, audio or video content that resembles an existing person, object, place, entity or event and could falsely appear authentic or truthful. An intention to deceive is not the sole consideration. The assessment also reflects the level of resemblance, the message conveyed, the deployment context and the expectations of the audience.
A wholly fictional avatar that does not resemble an existing individual is therefore not automatically a deepfake under the statutory definition. Disclosure may nevertheless be required under consumer law or advertising self-regulatory rules where the avatar is presented as a genuine customer, expert or witness and is capable of misleading the audience.
The notice must be perceivable upon first exposure
Technical marking embedded by the provider is not, in itself, sufficient to satisfy the obligation to inform the public that may apply to the deployer. The European Commission’s transparency guidance requires clear, distinguishable and accessible disclosure no later than first exposure. An image may carry a visible notice, an audio advertisement may use an audible announcement, and a video may combine an icon with explanatory text.
The wording should identify the relevant element, for example: “ The character featured in this advertisement was generated using artificial intelligence.” A generic notice hidden in terms of use or shown only at the end of a sequence may be inadequate. For evidently artistic, creative, satirical or fictional works, the manner of disclosure may be adapted so that it does not interfere with the work, but the duty is not simply removed.
Why is labelling insufficient to secure a campaign?
Advertising claims must remain accurate and substantiated
Article L. 121-2 of the French Consumer Code prohibits false or misleading statements and presentations relating to a product’s essential characteristics, expected results, origin or the advertiser’s commitments. The rule applies regardless of the technology used to produce the visual.
Artificially flawless skin, a simulated product demonstration, a fabricated manufacturing environment or a synthetic testimonial may distort consumer perception. An “AI-generated” notice does not cure an unsupported commercial promise. Marketing teams must be able to substantiate performance claims and distinguish creative illustration from evidence of the product’s actual effect.
To learn more about this transparency obligation, we invite you to consult our previously published article: “What legal challenges arise from the use of artificial intelligence in advertising content by influencers?”
Third-party rights require a separate clearance review
Indicating that content has been generated or modified using AI does not remove the need to ensure that its use respects third-party rights. Article L. 122-4 of the French Intellectual Property Code prohibits the unauthorised reproduction, adaptation or transformation of protected works. An advertising campaign may therefore reproduce or transform a protected work without authorization, but may also infringe trademark, design right, image or voice right, or damage an individual’s reputation.
Where prompts or uploaded files contain photographs, voices, customer profiles or internal documents, the GDPR and confidentiality obligations must also be considered. The CNIL recommends governance involving legal, operational, security and data-protection teams. A closed solution, or one configured to prevent provider reuse of data, may be preferable for sensitive campaign assets.
To learn more regarding the protection of rights, we invite you to consult our previously published article: ” How can one secure or assign rights in a work created with the assistance of artificial intelligence?“.
How should compliance for AI advertising be organised?
Implement a four-stage approval process
We recommend a process proportionate to the campaign’s visibility, budget, duration and territories:
- Document the tools, prompts, source files, successive versions and human modifications to better manage the risks associated with the use of artificial intelligence.
- Identify the provider, deployer, responsible advertiser and contractors involved in distribution.
- Clear rights in works, trademarks, music, voices, individuals and data used or reproduced.
- Determine before release whether the content requires technical marking, visible or audible disclosure, modification or regeneration.
Secure the entire contractual chain
Agreements with agencies, studios, influencers and technology providers should require prior notice of AI use, define approved tools, protect confidential material, require delivery of prompts and versions where appropriate, address intellectual property warranties and allocate claims management. A platform’s permission for commercial use does not guarantee that the output is free from third-party rights.
The approval file should remain available after release. Penalties for breach of the transparency requirements may reach EUR 15 million or 3% of total worldwide annual turnover, without excluding takedown measures, misleading-advertising proceedings or compensation claims by right holders.
Conclusion
The use of AI in advertising does not create a general duty to disclose every algorithmic intervention. It does require a structured assessment of realism, apparent authenticity and the likelihood of confusion. Where content constitutes a deepfake, disclosure must be clear and immediate. In every case, the advertiser must still review the accuracy of the message, third-party rights, the data used and the commitments of its contractors.
Dreyfus Law Firm assists its clients in managing complex intellectual property cases, offering personalized advice and comprehensive operational support for the complete protection of intellectual property.
Dreyfus Law Firm works in partnership with a global network of attorneys specializing in Intellectual Property.
Q&A
Who is responsible for labeling an AI-generated advertisement: the advertiser, the agency or the platform?
It depends on the role of each party in the campaign. The obligation to inform the public generally falls on the party using the AI system and disseminating the relevant content. Contracts between the advertiser, agency and platform may allocate responsibilities, but they do not exempt each party from complying with the obligations applicable to it.
What happens to an advertisement created before August 2, 2026 but first released after that date?
The creation date alone does not remove the duty. Content that had not already been made available to the public before August 2, 2026 should be assessed under the rules applicable when it is released. The limited transitional arrangements mainly concern certain provider marking obligations and do not create a general exemption for campaigns prepared earlier.
Can a platform require broader AI labelling than the AI Act?
Yes. A social network, advertising platform or an advertising network may impose additional contractual requirements through its terms or advertising policies. Advertisers must comply with both applicable law and the distribution channel’s rules, or risk removal or suspension of the campaign.
Must an AI disclosure be translated for every country in which the campaign appears?
The information must be understandable to the target audience. A multilingual campaign will therefore generally require translation or local adaptation. An icon can support the message, but it may not replace clear text or an audible statement in a language understood by the audience.
How long should prompts and campaign versions be retained?
There is no single retention period for all projects. Records should be kept for at least the exploitation period, the duration of contractual warranties and a reasonable period during which a claim may arise. International or reusable campaigns will generally justify longer retention than short-lived content.
This publication is intended for general public guidance and to highlight issues. It is not intended to apply to specific circumstances or to constitute legal advice.