Consent, Control and Compensation
Jollywood
Rights holders aren’t against AI. They’re against losing control.
In industry discussions about AI-generated content featuring someone’s IP, the objection rights holders raise is rarely a flat no. It’s closer to: not like this.
The objection isn’t to the technology. It’s to content made with their characters, their faces and their brands, without anyone asking, under terms nobody agreed to, with nothing coming back to them. That’s why the language the creative industries used in their recent court filings keeps returning to the same three ideas: consent, control and compensation.
Those three words are easy to say and harder to deliver. Each one breaks down into specific things a rights holder needs to be able to do. This piece looks at what they are, why the current market mostly can’t provide them, and what it would take.
Consent: opt in, not opt out
Consent sounds simple. In practice it has three layers.
First, the default has to be no. IP shouldn’t be available for AI creation until the rights holder says it is. The models that are working point the same way: Warner Music’s licensed AI music deal with Suno is opt-in, and ElevenLabs’ Iconic Marketplace has rights holders approve each request before anything is made.
Second, consent has to be specific. Agreeing to one kind of use is not agreeing to every kind. A rights holder may be glad to see their character in fan-made shorts and want nothing to do with political content or adult themes. Consent that can’t tell those apart isn’t really consent.
Third, the person is not the same as the property. Characters and brands are owned by companies. A likeness or a voice belongs to a person, and that person’s rights don’t transfer just because a studio owns the film they appeared in. The Disney and OpenAI agreement announced in December 2025 licensed more than 200 characters but excluded talent likenesses and voices. That line exists for a reason. Likeness is the most personal right in this market, and it needs the most explicit consent.
The NO FAKES Act, which cleared the Senate Judiciary Committee unanimously in June 2026, would create a federal right over a person’s voice and visual likeness. Whether or not it becomes law in its current form, the direction is clear: using someone’s likeness will require their permission, and that permission will need to be provable.
Control: the terms have to hold every time
Consent answers whether a use is allowed. Control is about how that answer is enforced across thousands of individual creations, most of which the rights holder will never personally see.
In traditional licensing, control comes from approvals. A licensee submits a product, the licensor signs off or sends it back. That works when there are dozens of submissions a year. It doesn’t work when there could be thousands a day. So control in AI creation has to look different. It needs four things.
- Terms that apply automatically. The rights holder sets the rules once: permitted uses, prohibited themes, contexts, formats. Every request is checked against them before anything is generated.
- Human review where it matters. Some uses are routine. Others deserve a person’s judgment. The rights holder should decide which is which, and the sensitive ones should come to them for approval.
- Hard limits. Some categories should never be possible, regardless of who asks or what they offer to pay.
- The right to change your mind. A rights holder’s comfort level will shift as they see how their IP is used. They need to be able to tighten terms, pause availability or withdraw entirely, and have that take effect going forward.
There’s also a quieter form of control that matters just as much: independence from any single platform. When OpenAI shut Sora down in March 2026, Disney’s character license on that platform ended with it. A rights holder whose terms only exist inside one company’s product doesn’t fully control them. Control means the terms belong to the rights holder and work wherever authorized creation happens.
Compensation: paid on use, with the receipts
The deals making headlines are priced like acquisitions. In August 2026, the Los Angeles Times reported that Google had approached major studios about licensing their IP for AI, with one person familiar estimating around $40 million per character on average. Numbers like that are real money for the largest catalogs. They also make clear who that model is built for.
Most rights holders are not Disney. They’re independent creators with a character people love, mid-sized brands, estates, athletes, actors, and the agencies that represent them. For them, compensation can’t depend on landing a nine-figure deal with a model company. It has to work at the level of individual uses.
That means three things:
- Payment tied to use. When licensed IP is used, the rights holder earns. More use, more revenue.
- Visibility. Rights holders should be able to see how their IP is being used: what’s being made, how often, and in what contexts. Without that, they’re being asked to trust a number.
- Proof. Each authorized piece of content should be traceable back to the license that allowed it. That protects the rights holder and the creator, because it separates legitimate work from the unauthorized copies that will always exist alongside it.
Compensation is also what changes the relationship between rights holders and AI. Today, most rights holders only experience AI as a source of unauthorized use to chase down. When authorized use pays, the same creative demand that currently costs them becomes a revenue line.
One character, three requests
Here’s how the three principles work together, using one of the characters on our platform. Fusillia is owned by SpaghettiHeadz, a hair accessory company. Say SpaghettiHeadz has made Fusillia available for AI creation and set its terms once: creator videos and fan content are allowed, brand partnerships need the company’s approval, and political, adult and gambling content are off limits. So are health and nutrition claims, a restriction specific to this brand that no standard list would include.
An approved use. A creator wants to make a short video of Fusillia dancing through a colorful city and post it to their public channels. That falls inside the terms, so the creator licenses it and makes the video. SpaghettiHeadz is paid when the license is bought and earns more as the video is viewed. The video carries a record of the license that allowed it.
A blocked use. Another creator asks for Fusillia promoting a diet supplement. Nothing about that is obviously offensive, but it crosses the health and nutrition line SpaghettiHeadz drew for its own brand, so the request is stopped before anything is generated. Nothing is made, so nothing needs to be taken down later.
A use that needs a person’s judgment. A fashion retailer wants Fusillia in a back-to-school campaign. The terms say brand work needs approval, so the draft goes to SpaghettiHeadz for review. If they approve, the retailer gets its video under SpaghettiHeadz’s terms, and SpaghettiHeadz is paid for it. If they decline, it isn’t released.
In each case SpaghettiHeadz set the rules once. Consent, control and compensation then happened request by request, without a new negotiation.
Where agencies and representatives fit
There’s a version of this story where automated licensing cuts out the people who negotiate deals today. We think that version is wrong.
Agencies, licensing agents and managers bring things software can’t: judgment about what a client should and shouldn’t do, relationships across the industry, and leverage in negotiation. Those don’t stop mattering because the volume of potential uses went up. If anything, they matter more, because a mistake now repeats at scale.
What changes is where the representative’s work lands. Instead of negotiating every individual use, they negotiate the framework: the approved uses, the off-limits categories, the pricing, the approval rules, the exclusivity. Infrastructure then applies those terms to every request automatically and reports back on what happened.
The representative keeps their role. The client keeps control. The paperwork that made small uses uneconomical goes away.
Union and guild agreements belong in the same picture. Contracts negotiated by SAG-AFTRA, the Writers Guild and others in recent years include AI protections, and those obligations don’t disappear in an automated system. The right approach is to build them into the terms, not around them.
What this adds up to
Consent, control and compensation aren’t three separate features. They’re one system seen from three sides. Consent without control is a promise nobody can keep. Control without compensation is just a better way to say no. Compensation without consent is the problem the industry is already fighting.
Our view is that all three have to be designed together, as one system, from the start. Bolting consent onto a platform built for scale, or compensation onto one built for control, leaves a gap someone will eventually fall through. That principle is behind Jollywood’s proprietary, patent-pending rights infrastructure.
Jollywood is the commercial rights infrastructure for AI-generated media. It’s in private beta, with applications open for rights holders, creators and partners at jollywood.ai.
Sources
- Disney and OpenAI agreement: Disney press release
- Sora shutdown and end of the Disney deal: Variety
- Google studio talks and per-character estimate: Los Angeles Times, syndicated
- Warner Music and Suno: Reuters via The Star
- ElevenLabs Iconic Marketplace: ElevenLabs
- NO FAKES Act: Congress.gov
