Rights, Ethics & Disclosure Primer

Not legal advice. This is general information to help you ask the right questions. Laws differ between countries and are changing fast, and platform terms change without notice. For commercial, client or distribution work, read the actual license text and consult a qualified entertainment or IP lawyer.

The five questions to ask about every asset #

1. Can I use this commercially? Check the model license and platform terms, and which plan you're on.
2. Who owns the output? Your plan's terms, plus copyright law in your country.
3. Is there a real person in it? Likeness or voice means you need consent.
4. Does it copy someone else's work? Characters, logos, music, recognizable art.
5. Have I disclosed how it was made? Platform, festival and legal requirements.

Track the answers in the Asset & rights tracker (Pre-Production Kit §10).

1. Model licenses and platform terms #

"Free to download" doesn't mean "free to use commercially." Licenses in this field are unusually varied.

Things to check:

• Does a free tier allow commercial use? Often it doesn't.
• Does cancelling your subscription affect rights to outputs made while subscribed?
• Are there content restrictions (no real people, no political content, etc.)?
• Is attribution required?
• LoRAs and fine-tunes carry their own licenses, and generally inherit the base model's restrictions. A Civitai LoRA marked "no commercial use" stays non-commercial even on a permissive base model.
• Workflows that chain several models are limited by the most restrictive license in the chain.

Training-data provenance: some productions, especially brand and client work, need models trained on licensed data so they can answer provenance questions. Some vendors market models on exactly this basis. Ask whether your client requires it.

2. Copyright in AI-assisted work #

Can you own what you make? #

In the United States, the Copyright Office's position (in its Copyright and Artificial Intelligence reports and registration guidance) is broadly:

• Material generated purely by AI from prompts isn't protected by copyright. Prompts alone are generally not treated as enough human control over the output.
• Human contributions can be protected: your script, your own footage and artwork, your selection and arrangement of AI material (e.g. the edit), and meaningful modifications you make to AI output.
• When registering, you must disclose AI-generated material that is more than minimal, and describe the human contribution.

What this means in practice: the more of you in the work (writing, editing, compositing, sound design, performance), the stronger your ownership claim. Document your process, because your breakdown is also evidence of human authorship.

Other countries differ. For example, the UK has a provision for "computer-generated works," and courts and offices elsewhere have reached varying conclusions. Check where you'll distribute.

Infringement: copying others #

AI output can still infringe someone else's copyright or trademark. Avoid:

• Copyrighted characters (even "in a different style")
• Recognizable artworks, posters or album covers
• Logos and trademarks, which often appear by accident on clothing, signs and products, so check every frame
• Prompts that name living artists to imitate their specific style. This is legally unsettled and ethically fraught; many communities and festivals discourage it.
• Close copies of training data, such as a near-duplicate of a famous photo

AiFC's Challenge Notice asks every entrant to respect third-party IP, and a challenge brief can add its own rules on characters and music.

3. People: likeness, voice and consent #

The rule #

Don't depict a real, identifiable person, or clone their voice, without their informed, written consent. This includes celebrities, public figures, friends, family and yourself in contexts you didn't intend.

Why #

• Right of publicity / personality rights: many jurisdictions give people control over commercial use of their name, image and voice.
• Deepfake and digital-replica laws: a growing number of laws target non-consensual synthetic depictions, especially sexual content and election-related content.
• Platform rules: most generation platforms and distribution sites prohibit non-consensual real-person content.
• Ethics: it can cause real harm to real people.

When you do use a real person (with consent) #

• Get written consent specifying what you'll make, where it will be shown, for how long, and whether it can be commercial.
• For voice cloning, many voice platforms now require verified consent from the voice owner. Keep the record.
• For professional performers, check union rules and guidance (e.g. SAG-AFTRA's digital-replica provisions).
• Keep consent forms with your project files.

Synthetic characters #

Designed characters and voices that don't resemble real people are fine. Do check that your "original" character hasn't drifted into looking like a specific celebrity.

4. Music and sound #

Music is the most common rights problem in finished films.

Sound effects follow the same logic: check each library's license.

Content ID: platform matching systems can flag your film even if you hold a license. Keep your license documentation handy for disputes.

5. Disclosure #

Why disclose #

• It's increasingly required by platforms, festivals and, in some places, law.
• It builds trust. Clear disclosure of tools and process reads as confidence.
• It's AiFC's whole philosophy: process over prestige.

Platform rules (October 2026 snapshot, so check current policy) #

• YouTube requires creators to disclose realistic altered or synthetic content at upload. From May 2026 it can apply labels automatically when it detects significant photorealistic AI use and the creator hasn't disclosed. Labels are prominent for photorealistic content and appear in the description for clearly unrealistic or animated content. YouTube says a label alone doesn't affect monetization or recommendations, and content carrying C2PA metadata indicating full AI generation is labeled automatically.
• Other platforms (TikTok, Meta, etc.) have their own AI-labeling rules and auto-detection. Check each one before posting.

Law (examples) #

• EU AI Act, Article 50: transparency obligations apply from August 2, 2026. Providers of generative systems must mark outputs in a machine-readable way, and deployers must clearly label deepfakes. A Code of Practice on marking and labeling (published June 2026) describes layered marking (metadata plus watermarks) and labels such as an EU icon. Read the official guidance on how it applies to artistic, creative and satirical works.
• Other jurisdictions have or are adding similar rules, especially for political content and non-consensual imagery.

Festivals and competitions #

• Many AI festivals require a breakdown or disclosure. Some traditional festivals restrict or exclude AI work, or have separate categories.
• Read the eligibility rules before you start, especially premiere requirements and "percentage AI" rules.

Content Credentials (C2PA) #

C2PA is an open standard for attaching tamper-evident provenance metadata to media: which tools were used, and what was edited. Some generation and editing tools add it automatically. Platforms can read it, and anyone can check a file at contentcredentials.org. Consider keeping it on your exports as part of honest disclosure.

AiFC disclosure standard #

AiFC asks every challenge submission to include, in its write-up (see the Process Breakdown Template):

• The tools used
• The approximate AI vs. hand-made split, and what was hand-made
• Rights confirmation for music, voice and likeness

6. Ethics beyond the law #

Legal isn't always the same as right. Community norms on AiFC:

• Credit people. Name workflow authors, LoRA creators, collaborators, voice actors and asset sources.
• Don't pass off AI work as something else, such as claiming generated footage was shot practically.
• Be careful with styles. Draw on influences broadly rather than cloning a living artist's signature look.
• No harmful content: no non-consensual intimate imagery, no harassment, no deceptive political deepfakes, nothing sexualizing minors, ever.
• Consider the crew you're replacing. Many AI filmmakers collaborate with human actors, voice artists, composers and editors. Those collaborations often make better films and a healthier industry.
• Be honest about limits in breakdowns. Cherry-picking is normal; claiming one-shot perfection isn't.

7. Pre-release checklist #

• Every model, LoRA and platform used allows my intended use (personal, monetized, client, festival)
• No real person's likeness or voice without written consent on file
• No copyrighted characters, logos, artworks or trademarks visible (checked frame by frame)
• Music and SFX licensed for this use, with attribution added where required
• AI disclosure written (breakdown, platform label, festival form)
• Content Credentials kept on the export (optional, recommended)
• Festival or competition eligibility rules read (AI rules, premiere status, runtime)
• For client or commercial work: license texts saved, and a lawyer consulted if anything is unclear

Sources & further reading #

• U.S. Copyright Office: Copyright and Artificial Intelligence
• USCO letter on the AI and Copyright Initiative
• Assessing Human Authorship in AI-Generated Works (NYSBA)
• EU AI Act Article 50 explained (LearnSignal)
• European Commission final Code of Practice on marking and labelling AI-generated content (Jones Day)
• YouTube auto-labels undisclosed AI videos (TechWyse)
• Content Credentials / C2PA
• SAG-AFTRA (search "artificial intelligence" for digital-replica guidance)
• Black Forest Labs licensing