Picture the Monday after a wedding in Fort Lauderdale. The gallery lands, the couple posts their favorite frame to Instagram, and a small tag appears under the caption: AI info. Nobody faked anything. The retoucher used a generative brush to remove an exit sign from the ballroom wall. That single stroke wrote a provenance record into the file, and the platform read it.

That is the modern version of the question in the title. Not "is this photo fake," but "who decided what counts as an edit, who has to disclose it, and who owns the result." Adobe's Creative Cloud apps embed C2PA Content Credentials — a cryptographically signed record of the tool used and the edits applied — and Meta has said it labels images on Facebook, Instagram and Threads when it detects the "AI generated" markers in the C2PA and IPTC standards. The file talks. Your contract should decide what it says.

If you are booking photography and videography in South Florida this year, the paragraphs below are the ones worth arguing over before you sign — not the deposit schedule.

Define the edit line, because "retouching" no longer means one thing

Every generative editing conversation collapses without a shared vocabulary. A contract that says "standard retouching included" was written for a world where retouching meant dodging, burning and skin work. It now silently authorizes a photographer to rebuild half a frame.

Ask for three named tiers, with the middle one requiring your written sign-off:

The reconstruction tier is where families get hurt. Composite group shots and "opened" eyes are the ones that show up years later as a quiet argument about whether a moment happened. Reasonable people disagree — some couples want the fix, some are horrified by it. The contract's job is to make sure the couple decides, not the editor at 2 a.m.

One practical add-on: ask whether Content Credentials will be preserved, stripped or disclosed on delivered files, and whether the gallery host preserves them. If the answer is "I don't know," that is your cue to keep asking.

Ownership gets stranger when a machine did part of the work

Most event contracts hand the client a broad usage license while the photographer keeps copyright. That structure still works — but generative content sits outside it in a way most templates never anticipated.

The U.S. Copyright Office's position is blunt: human authorship is the bedrock requirement, and material generated entirely by an AI system is not protected. Its Part 2 report on copyrightability, published January 29, 2025, concluded that using AI as an assistive tool does not strip protection from a human-authored work, while purely prompt-generated output is not registrable. Where a work mixes both, only the human contribution is protectable.

There is a second, more operational consequence. Under the Office's registration guidance, applicants have a duty to disclose AI-generated content in a work submitted for registration and to describe the human author's contribution. For a corporate brand film or a documentary-style event piece that someone intends to register, undisclosed generative content is a defect in the filing, not a stylistic footnote.

So the ownership clause needs one added sentence: the deliverables contain no generative material other than what is itemized in Exhibit A, and any such material is identified with the tool used. That single line converts a future dispute into a document.

AI training image rights: the clause almost no template has

Your wedding gallery, your keynote footage, your product-launch stills — every one of them is a labeled dataset. Faces, venue interiors, food styling, brand marks, all neatly organized by event. Nothing in a traditional contract stops a vendor, a gallery platform or a downstream editing service from licensing that library for model training.

Ask for an explicit prohibition, and make it travel:

• No training, no exception. The vendor will not use, license or permit the use of your images, footage, audio or derived embeddings to train, fine-tune or evaluate any AI model.
• Flow-down. The same restriction binds second shooters, editors, colorists, retouching houses, gallery hosts and cloud storage.
• Named tools. The vendor discloses which AI-enabled software touches your files, and the terms governing that software. Major vendors now publish explicit generative AI terms — Adobe's generative AI user guidelines are one example of the kind of document you are entitled to see.
• Signal it in the file. The C2PA specification supports assertions covering training and data mining permissions that ride along with the asset. A contract clause protects you against your vendor; a "do not train" assertion travels with the file after it leaves everyone's hands.

Corporate clients should treat this as a data-governance item, not a creative one. It belongs in the same review as your NDA. Our corporate events clients routinely have legal or comms teams who want media rights aligned with company policy — bring them in before the contract is signed, not after the footage is delivered.

Likeness, guests and the line that turns a photo into an advertisement

Florida already gives every person a statutory claim here. Fla. Stat. § 540.08 prohibits publishing a person's name, portrait, photograph or other likeness for trade, commercial or advertising purposes without express written or oral consent, and lets that person sue to stop the use and recover damages, including a reasonable royalty. Generative AI does not create that risk; it multiplies the number of ways to trip over it.

Three concrete scenarios worth naming in the contract:

1. The vendor's own marketing. A portfolio-use clause is normal. A portfolio-use clause that permits generative alteration of guests' faces or bodies is not. Limit vendor promotional use to unaltered or correction-tier images, and give guests a removal request path.
2. Synthetic guests and synthetic crowds. Adding people who were not there, or making a half-full room look packed, moves an image from documentation to representation. For corporate work that becomes a recruiting or sales asset, that is a claims problem.
3. Digital replicas. Federal law is moving. The Copyright Office recommended a federal digital replica statute in Part 1 of its AI report, and the NO FAKES Act of 2026 would create a licensable federal property right in an individual's voice and visual likeness, with a notice-and-takedown process. It advanced out of the Senate Judiciary Committee in June 2026 and is not law yet — but a contract drafted today will still be governing your files when it is.

If your company has an EU parent, subsidiary or audience, add one more: the EU AI Act's transparency rules under Article 50 apply from 2 August 2026, requiring machine-readable marking of synthetic content by providers and disclosure of deepfakes by deployers. Whoever publishes the video carries the disclosure duty, not whoever edited it.

Where AI turns your event media into a legal claim

Event photography rarely lives alone. It ends up in case studies, testimonial reels and paid social. That is where a separate rulebook applies.

The FTC's Rule on the Use of Consumer Reviews and Testimonials (16 CFR Part 465) took effect October 21, 2024. It covers reviews and testimonials that misrepresent being from someone who does not exist — the FTC names AI-generated fake reviews specifically — or who had no actual experience with the business, and prohibits businesses from creating, selling or disseminating them when they knew or should have known they were false.

Practical translation for anyone producing event content: a synthetic on-camera guest praising a venue is not a creative choice, it is a prohibited testimonial. An AI voiceover reading a real client's words is fine if the words and the person are real and the presentation is not misleading. A composited "attendee reaction" is not. Build that distinction into the media-release language you collect at the door, and into whatever your event planning and contract expectations workflow hands to the production team.

The eight clauses to request, in plain language

• Edit tier definition. Correction, reconstruction and generation defined by example, with approval rights attached to each.
• Disclosure. A written list of AI tools used on your files, delivered with the gallery.
• No training. Prohibition on training, fine-tuning or dataset licensing, binding all subcontractors and platforms.
• Provenance handling. Whether Content Credentials are preserved or stripped, and who decides.
• Likeness limits. No generative alteration of any identifiable person in vendor marketing; guest removal on request.
• Event video usage rights. Term, territory, media, edit rights and whether AI reframing, upscaling or dubbing is permitted — separately from the license grant itself.
• Archive and deletion. How long raw files are retained, where, and what a deletion request actually deletes.
• Authenticity warranty. The vendor warrants the deliverables document the event as it occurred, except as itemized.

Photo contract: what to ask before you sign

1. Which AI features are enabled in your editing workflow by default, and can they be turned off for my files?
2. What is your policy on composite group shots and opened eyes — do you do it, and do you tell me when you have?
3. Do the delivered files carry Content Credentials, and does your gallery host preserve them?
4. Will you sign a no-training clause that binds your editors and your storage provider?
5. If I request an unedited or correction-only version of a specific image, is that included or billed?
6. Who holds copyright, and how are any generative elements identified if I need to register the work?
7. What are the exact usage rights for the video — including whether AI reframing or translation counts as an authorized edit?
8. How long do you keep raw footage, and what happens to it if you close the business or sell it?

Vendors who work with AI thoughtfully answer these quickly. Hesitation on question four is the most informative answer you will get all meeting.

FAQ

Is AI editing on wedding photos a bad thing?

No. Noise reduction, masking and subject selection are AI features that make good photographs better and have been standard for years. The issue is not the technology, it is the absence of a defined threshold between fixing an image and rewriting one.

Why did my real photo get labeled as AI on Instagram?

Because platforms read provenance metadata rather than judging the picture. Editing software can write a Content Credential when any AI-assisted feature is used, and Meta has said it applies labels when it detects the standard AI markers. A minor generative cleanup and a fully synthetic image can produce the same signal.

Can I own a photo that was partly generated by AI?

Copyright protects the human-authored contributions. The Copyright Office has said material generated entirely by AI is not protectable, and that applicants registering a work must disclose AI-generated content and describe the human contribution. For most event photography the human authorship is overwhelming — but the disclosure obligation is real if you register.

Can my photographer sell my wedding images to an AI company?

Unless your contract prohibits it, the question is governed by whatever the vendor's terms and their platforms' terms say. That is precisely why the training clause needs to be written in rather than assumed.

Do these rules apply to a corporate event as well as a wedding?

The commercial-use exposure is usually higher. Corporate footage becomes advertising, recruiting material and investor collateral, which pulls in likeness law, advertising rules and, for EU-facing organizations, AI Act transparency duties.

What if the contract is already signed?

Most vendors will accept a short written addendum covering training, edit tiers and disclosure. It is a far easier conversation before delivery than after.

A necessary caveat

This article is general information about how contract language is evolving, not legal advice, and it does not create an attorney-client relationship. Statutes, agency guidance and pending legislation cited here change; the NO FAKES Act in particular has not been enacted. For contracts that carry real financial or reputational exposure, have a Florida-licensed attorney review the final language.

Conclusion: decide now who gets to change the moment

Every photo and video contract already answers this question. Most of them answer it by silence, which means the answer is "whoever is holding the file." Ten years from now, the images from your wedding or your company's anniversary will still exist, still be readable by systems that did not exist when they were made, and still carry whatever provenance record was written into them the week they were edited.

You get exactly one cheap opportunity to set the terms: the hour before you sign. Define the edit line. Kill the training rights. Ask what the metadata will say. Then let the photographer do what a photographer is actually for — being in the right place when something real happens.