The release review

This is operational guidance, not legal advice. Requirements vary by country, audience, product category, platform, contract, and the details of the generation.

Current legal and policy signals

• Truth in advertising: The FTC says advertising claims must be truthful, not deceptive or unfair, and evidence-based. AI does not create an exemption.10
• Endorsements and testimonials: The FTC’s endorsement guidance applies across media. An endorser should not claim product experience they did not have or make a claim the advertiser cannot substantiate. The FTC’s 2024 fake-review rule addresses fake consumer, celebrity, and AI-generated testimonials.2425
• Impersonation: The FTC has proposed and pursued tools against AI-enabled impersonation and warns that voice cloning can turbocharge fraud. A commercial digital double needs explicit identity rights and a non-deceptive presentation.26
• EU transparency: The European Commission says Article 50 transparency rules apply from August 2, 2026; chatbots must tell people they are dealing with AI, and deepfakes and some AI-generated/altered content require labeling and machine-readable marks.9
• Copyrightability: The U.S. Copyright Office’s Part 2 report says AI-assisted works can receive protection where a human determines sufficient expressive elements; prompts alone are not enough. Its Part 3 work on training data remains a separate issue.1427
• Provenance: C2PA Content Credentials can record signed provenance, edits, and AI use. They help people inspect origin but do not prove the output is lawful or true and should not be the only disclosure mechanism.11
• EU implementation detail: Article 50 requires providers of systems that generate synthetic audio, image, video, or text to make outputs machine-readable where technically feasible, and requires deployers to disclose deepfakes. Artistic, satirical, or fictional work still needs disclosure in an appropriate way when the audience could be misled. The Commission says the transparency rules apply from August 2, 2026; enforcement is primarily through national authorities. See the Commission FAQ and the consolidated AI Act text.
• U.S. synthetic testimonials: The FTC’s final rule, effective October 21, 2024, addresses fake or false consumer, celebrity, and AI-generated reviews and testimonials, as well as fake social indicators. A fictional actor playing a role is not automatically safe: the underlying ad can still mislead if viewers are led to believe the person is real, independent, or an actual user. See the FTC final rule announcement and Q&A.
• Voice and impersonation: The FTC says there is no AI exemption from existing law and highlights voice-cloning risks. Treat a voice as a rights-bearing identity signal, not merely an audio texture. See FTC voice-cloning guidance and FTC AI claims guidance.
• UK advertising practice: The ASA/CAP position is media-neutral: the same rules apply whether an ad was generated by a person, a model, or both. Their 2025 guidance emphasizes disclosure where omission could mislead and warns advertisers to check bias, harmful stereotypes, misleading images, endorsements, and testimonials. See ASA disclosure guidance and ASA policy monitoring.
• Platform labels are not a legal safe harbor: Google’s 2026 documentation describes “How this ad was made” transparency and direct labels for certain AI-generated or modified creatives; Meta describes “AI info” in “About this ad”; TikTok’s Symphony materials describe AI-generated labels and advertiser responsibility. The label is one control in a broader review, not permission to use an unlicensed face, false claim, or hidden targeting rule. See Google’s transparency update, Meta’s ad transparency update, and TikTok Symphony.

Rights matrix

A campaign release checklist

Brief: named business owner, creative owner, legal owner, audience, claim list, prohibited use list, disclosure plan.

Inputs: rights to reference images, footage, music, voices, faces, data, copy, and brand assets.

Generation: model/provider, plan, date, prompt/input, seed or reference when available, output ID, operator.

Review: factual QA, product QA, cultural review, accessibility, likeness, bias, brand safety, platform policy, disclosure, local law.

Publication: final file hash/version, C2PA or other provenance where available, caption/label, landing page, audience settings, media placement, monitoring owner.

Aftercare: correction path, takedown path, model/vendor change alert, incident log, performance review, asset expiration.

Red lines

• Never invent a testimonial, customer experience, product result, medical/financial outcome, award, statistic, or quote.
• Never clone a real person’s face or voice without documented consent and a lawful, scoped use.
• Never use a stock avatar in paid advertising when the provider’s license excludes that use.
• Never treat a model’s “commercial use” statement as clearance for the prompt inputs or third-party references.
• Never let a platform silently generate or modify a paid ad without an owner checking what is actually delivered.
• Never use personalization to infer or target sensitive traits without a documented legal and ethical basis.
• Never use AI disclosure as a substitute for truthfulness; a clearly labeled false ad is still false.