AI Content Ownership: Who Owns What When AI Makes Your Ad?

You typed a prompt, the model spat out a slick 15-second video, and now it's running as a paid ad. Simple question: who owns it? The honest answer is that "own" is doing three jobs at once, and most people only think about one of them.

There's the question of whether you're allowed to use the ad at all. There's the question of whether you can stop a competitor from ripping it off. And there's the question of who gets sued if the ad turns out to infringe someone else's work. These are three different problems with three different answers, and the platform's terms of service only settles one of them.

Get these confused and you'll either panic about a risk that doesn't matter or ignore the one that does. Let me untangle them.

The three questions hiding inside "who owns it?"

Ownership and protection sound like the same thing. They're not. A tool can hand you full commercial ownership of an output that copyright law refuses to protect. You'd have something you're allowed to run everywhere but can't stop anyone from copying. That's not a contradiction, it's just how the two systems work.

For most advertisers, the ranking of these questions is backwards from what they assume. Protection is usually the least important. Liability is usually the most.

Question 1: Can you use it? (Ownership)

This one is straightforward and it lives entirely in the terms of service of whatever tool made the video. Most major generators now assign output rights to the user, but the details differ and the free tier is where people get burned.

A quick lay of the land based on current platform terms:

• OpenAI (including Sora) uses a rights-transfer model: it "assigns to you all its right, title, and interest in and to Output," which gives full commercial use on paid plans.
• Runway grants commercial rights on every plan, including free, and you keep ownership of your outputs. (source)
• Midjourney gives ownership to paid subscribers. On the free tier you don't own what you make, which matters if a freelancer quietly used a free account to cut costs.

The trap is the free-tier gap. A creator delivers you a batch of clips, you assume you own them, and it later turns out they were generated on a plan that never granted commercial rights in the first place. The tool's contract was with the creator, not you, and if the creator broke it, your "ownership" is built on sand.

Two rules that save you here:

1. Get commercial rights confirmed in writing from whoever delivers the asset, not from the platform's marketing page.
2. Know which tool and which plan tier produced the final file. "AI made it" is not an answer you can put in a contract.

Question 2: Can you stop copycats? (Protection)

Here's where the law gets blunt. In the US, purely AI-generated material cannot be copyrighted. The Copyright Office has been clear that copyright protects "only material that is the product of human creativity," and that when the AI "determines the expressive elements of its output, the generated material is not the product of human authorship." (source) Typing a prompt, however clever, doesn't count as authorship. The D.C. Circuit backed this up, and the human-authorship requirement now stands as settled law.

So if your ad was generated end to end from a prompt with no meaningful human shaping, you can't register it, and you can't easily stop a competitor from lifting it frame for frame.

Before you panic: for most performance ads, this barely matters. Nobody is building a moat around a 15-second TikTok creative. By the time a competitor could copy it, you've already tested it, learned from it, and moved on to the next batch. The value of an ad is in the performance data and the speed, not in owning the pixels forever. Copyright protection matters for evergreen brand films, mascots, jingles, and hero assets you plan to defend for years. It rarely matters for the disposable creative that makes up the bulk of a paid-media account.

The Copyright Office also left an important door open. AI assistance does not kill copyrightability. If a human "makes creative arrangements or modifications of the output," or a human-authored work is perceptible inside the AI output, the human contribution can be protected. Which brings us to the ladder.

The human contribution ladder

How much of your ad is actually protectable depends on how much a human shaped the expression, not just pressed generate. Roughly, from no protection to real protection:

1. Raw prompt-to-output, used as-is. Not protectable.
2. Generating dozens of variations and selecting one. Selection alone is weak, still basically not protectable.
3. Editing, sequencing, and arranging AI clips into a timeline with human creative choices. The arrangement and edit can be protected, even if individual clips aren't.
4. Combining AI elements with human-authored material: your script, your voiceover, filmed footage, original music, brand design. The human parts are protected.
5. AI used only as a background tool inside a human-directed production. Fully protectable as a normal creative work.

The higher you climb, the more of your ad you can defend. This is also why a skilled creator who directs, edits, and layers is producing something legally sturdier than a prompt jockey handing over raw generations. The human labor isn't just about quality, it's about what you can actually own.

Question 3: Who gets sued? (Liability)

This is the one that should keep an advertiser up at night, and it's the one platform ownership language quietly does not solve.

Owning an output doesn't mean the output is clean. If the model was trained on copyrighted material and your ad reproduces something recognizable, or if it generates a face that resembles a real person, you can face an infringement or right-of-publicity claim. Ownership says "you can use this." It says nothing about "this is safe to use."

That's what indemnification is for, and it's unevenly distributed:

• OpenAI offers IP indemnification for API and business or enterprise customers, covering certain third-party infringement claims. It does not extend that to consumer ChatGPT or Sora plans. (source)
• Runway, by contrast, has clean commercial terms but no indemnification if your output triggers a copyright claim.

Read that again: the tool with the friendliest ownership language can still leave you completely exposed if something goes wrong. Ownership and indemnification are separate promises, and the second one is worth more to a brand running ads at scale.

For a brand, the practical liability checklist is short:

• Does anyone indemnify you if the ad infringes? On which plan tier?
• Was any real person's likeness or voice used, and do you have a release?
• Does the output contain recognizable logos, characters, or copyrighted footage from training data?
• If a creator delivered it, does your agreement pass through their rights and warranties to you?

Who owns what in a real marketplace deal

The picture gets more interesting once you're not the one pressing generate. When a brand hires a creator who uses AI tools to make an ad, four parties have a claim on some slice of the outcome. Here's how it usually shakes out with a well-written agreement:

The failure mode is a handshake deal where none of this is written down. The brand assumes it owns everything forever. The creator assumes they can reuse the work in a portfolio or resell variations. The AI tool's free-tier restriction quietly voids the commercial rights everyone assumed existed. Nobody's lying, they just never aligned the three questions.

This is exactly why buying campaign-ready AI ads through a structured marketplace beats sourcing raw clips from a random prompt-for-hire. On Viralix, the rights, deliverables, and usage terms are part of the package, so the handoff from creator to brand isn't a guessing game. It's one less thing to litigate later. If you want the deeper mechanics of licensing terms and usage windows, our AI video rights and licensing field guide covers it.

The one-page decision rule

When someone hands you an AI-made ad, run it through this in order:

1. Can I use it? Confirm commercial rights in writing, and confirm which tool and tier produced the final file. If you can't, stop here.
2. Do I need to protect it? If it's a disposable performance ad, skip it. If it's a hero asset or brand mascot you'll defend for years, make sure a human shaped enough of the expression to climb the contribution ladder, and register the human-authored parts.
3. Who covers me if it's not safe? Check for indemnification, clear any likeness or third-party material, and make sure the creator's warranties flow through to you.

Most people obsess over step 2 because "can you copyright AI content?" is the question that gets all the headlines. For advertisers, steps 1 and 3 are where the real money and the real risk live.

Bottom line

AI-generated ads sit in a gap: you can usually own and use them, you often can't protect them, and you're frequently on the hook if they infringe. Owning an output is a contract. Protecting it is copyright, and it takes real human authorship to earn. Being safe to run it is indemnification, and it's the promise most tools quietly withhold.

Don't ask "who owns this?" Ask three questions instead: can I use it, do I need to protect it, and who pays if it goes wrong. Get those written down before the ad goes live, not after the takedown notice arrives.