The Bottom Line
- Beginning August 2, 2026, the EU AI Act will require disclosures for realistic AI-generated or AI-manipulated depictions of people, objects, places, entities or events that falsely appear authentic or truthful.
- This requirement also applies to AI-generated or AI-manipulated text used to inform the public on matters of public interest, including certain advertising, public relations and corporate communications; but meaningful human review and editorial responsibility eliminates the need for disclosure.
- Companies whose AI-generated advertising, marketing or PR content may reach EU audiences should take the following steps to prepare for compliance: (1) clearly label realistic AI-generated creative assets used in the market, including product shots, backgrounds, and human as well as non-human characters; and (2) clearly label AI-generated press releases and ad copy addressing matters of public interest, or else implement a robust human editorial review process.
On August 2, 2026, the EU AI Act’s transparency obligations under Article 50(4) take effect, requiring deployers of AI systems that generate or manipulate “deep fake” content to disclose its artificial nature. In anticipation of this deadline, the European Commission published final implementation guidelines (“Guidelines”) interpreting the scope and practical application of these disclosure obligations. The guidance clarifies that the term “deep fake” has a significantly broader meaning than is commonly understood in the U.S., and expands compliance obligations for brands, agencies and PR professionals using AI-generated content.
The AI Act’s Transparency Framework
The EU Artificial Intelligence Act, widely considered the world’s first comprehensive AI law, takes a risk-based, tiered approach in which AI systems presenting unacceptable risks are banned, higher-risk systems are subject to enhanced compliance obligations, and lower-risk systems face more limited requirements. The Act sets out requirements that apply to both AI providers and deployers, including organizations located outside the EU that use AI tools to produce output used within the EU.
Article 50 of the AI Act establishes a series of transparency and disclosure obligations that apply across risk tiers.
- Human interaction (Art. 50(1)): AI providers and deployers must ensure users are informed when they are interacting directly with AI systems such as chatbots and AI companions.
- Synthetic content marking (Art. 50(2)): Providers of AI systems must include machine-readable markers on AI-generated or AI-manipulated content (synthetic imagery, video, audio, or text) identifying it as artificial.
- Emotion recognition/biometric systems (Art. 50(3)): Deployers must inform individuals when these systems are being used.
- Deep fakes (Art. 50(4)): Deployers utilizing AI-generated or AI-manipulated deep fake content must disclose that such material is artificial. This requirement also extends to AI-generated text used to inform the public on matters of public interest, which may include certain types of advertising and public relations communications.
Article 50(5) further requires that all such disclosures be provided in a “clear and distinguishable manner,” at the latest at the time of first interaction or exposure, and consistent with applicable accessibility requirements. These transparency obligations take effect on August 2, 2026.
A Much Broader Definition of “Deep Fake”
The European Commission’s Guidelines provide important clarity on what constitutes a “deep fake” under Article 50(4). The AI Act defines a “deep fake” as an AI-generated or AI-manipulated image, audio, or video content that resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful.
This definition is significantly broader than the term’s common usage in the United States, where “deepfake” is generally associated with face-swapping, voice cloning or other attempts to create a deceptive digital replica of a real person. Under the AI Act, the term “deep fake” not only applies to such digital replicas, but also includes realistic AI-generated depictions of objects, places, animals, events and human avatars that do not resemble identifiable individuals. The focus is not whether a real person has been replicated, but whether the content appears authentic when it is, in fact, synthetic or materially manipulated.
The Guidelines also clarify that not all AI-generated content is considered a “deep fake.” Content that is not deceptively realistic, either because it is obviously fantastical, fictional or unrealistic, generally falls outside the definition because viewers would not perceive it as an authentic depiction of reality. Likewise, AI-generated content that is not material enough to impact a viewer’s assessment of authenticity does not constitute a deep fake.
For instance, routine editing and post-production activities utilizing AI technology, such as color correction, noise reduction, lighting adjustments or audio cleanup, typically will not trigger disclosure obligations where they do not affect the audience’s understanding of the content’s authenticity.
However, the use of AI tools for editing and post-production may result in a deep fake if the AI modifications impact essential elements of material that will affect the audience’s assessment of its contents, such as de-aging performers, or simulating a person’s performance or parts thereof.
Whether content is likely to appear deceptively authentic must be assessed in context, including consideration of the intended audience, because certain groups, such as children, older individuals or those with limited familiarity with AI-generated media, may be more susceptible to deception.
Creative and Artistic Content Carve-Out
Article 50(4) provides reduced disclosure requirements for deep fake content forming part of an evidently artistic, creative, satirical, fictional, or analogous work. Such AI disclosures must be provided in a manner that does not hamper the display or enjoyment of the artistic work.
Advertisers should approach this carve-out cautiously. The Guidelines make it clear that, where content combines both commercial and creative characteristics, the commercial character generally prevails. As a result, advertising, marketing and other commercial content will rarely qualify for this reduced disclosure treatment.
Examples Relevant to Advertisers
The Commission’s Guidelines provide helpful examples to determine what does and does not constitute a deep fake requiring a disclosure.
- Is a deep fake:
- An AI-generated image of a product in advertisement or packaging that can affect the audience’s perception and mislead as to the actual product appearance, characteristics or use (e.g. making the product appear not identical to the real product, more appealing or with improved quality than in real life).
- An AI-generated video featuring an AI-generated depiction of a celebrity influencer in an advertising or promotional context.
- An AI-manipulated video featuring a realistic synthetic influencer testing out a sponsored real product, focused solely on displaying the product functionalities.
- An AI-generated video featuring a scene involving two professional football players in front of a building resembling a football stadium.
- Is NOT a deep fake:
- An AI-generated video of mice arguing in human language over the best type of cheese, used as part of a cheese advertisement.
- A real product (e.g., a car) shown in an advertisement against an AI-generated background and surrounding environment, as long as the ad is not likely to mislead the audience about the product’s actual representation and its characteristics and use.
- An AI-manipulated radio broadcast adjusting technical audio parameters (such as volume levels, noise reduction, audio compression) without altering the actual words spoken by speakers or their way of speaking.
- An AI-generated cartoon of a pre-existing image depicting an historical event.
AI-Generated Text May Also Require Disclosure
The disclosure obligation is not limited to images, audio and video. Article 50(4) also requires disclosure when AI is used to generate or manipulate text intended to inform the public on matters of public interest. In advertising and PR, this requirement could apply to AI-generated product claims or publicity addressing health, safety, environmental, economic, financial, political, scientific or cultural issues, which may be considered subjects of public debate. Companies using AI to create or manipulate corporate reports or investor information will face the same disclosure requirement.
Notwithstanding the foregoing, disclosure is not required where the AI-generated text is subject to meaningful human review and editorial control, and a human editor or editorial entity (such as a publishing company) takes ultimate responsibility for the content. This exception is particularly important for advertisers and communications professionals incorporating AI tools into existing content creation workflows.
What Advertisers Should Do Now
With the August 2, 2026 compliance deadline approaching, advertisers, agencies and communications professionals should review their AI-generated creative assets, evaluate whether their campaigns may reach EU audiences and determine where disclosure mechanisms may be necessary. Organizations using AI to prepare public-facing written content should also assess whether existing editorial review processes are sufficient to satisfy the Act’s human review exception.
Why U.S.-Based Advertisers Should Pay Attention
Importantly, the impact of the EU AI Act extends beyond the borders of Europe. A U.S.-based advertiser, agency or other organization may fall within scope if its AI-generated content is used in the EU, even if it has no offices, employees or physical presence there. The Commission’s Guidelines specifically note that a non-EU advertiser using AI to generate a deep fake for an advertisement displayed in the EU may be considered a deployer under the AI Act.
As a result, Article 50(4)’s disclosure requirements may apply to a wide range of U.S.-based organizations whose AI-generated content reaches EU audiences, including:
- Global or pan-European digital advertising campaigns (e.g., programmatic display, video, or social media ads served to EU audiences);
- Social media advertising on platforms with EU users, particularly where campaigns target or are available to EU consumers; and
- EU-facing e-commerce websites or marketing landing pages featuring AI-generated product imagery, models, or scenes.
The key question is not where the advertiser is physically located or where the content was created, but whether the AI-generated output will reach EU audiences. Accordingly, advertisers should not assume that the absence of a European subsidiary or office insulates them from AI Act’s disclosure obligations.
This matters because current U.S. AI disclosure requirements remain comparatively narrow and fragmented. While certain state laws require disclosures for synthetic human performers, political advertising or specific AI applications, U.S. law generally does not impose a blanket disclosure requirement for realistic AI-generated depictions of products, places, animals or events. The EU framework reaches substantially further and does not require proof of deceptive intent. As a result, companies which may be compliant with applicable AI laws in the U.S., or which may not be subject to AI laws in the U.S., may face new compliance requirements in the EU.
Penalties
Companies that fail to comply with these new disclosure obligations beginning August 2, 2026 may be subject to a fine of as much as €15 million, or up to 3% of the company’s worldwide annual revenue for the previous fiscal year. The actual amount of the fine will depend on the size of the company and the relevant circumstances of the specific default, including its degree of seriousness, and any aggravating or mitigating circumstances.