AI, Culture, and Brand Protection: Who Actually Gets Paid

Culture / Ownership / Pablo Segarra

Picture a producer in a bedroom. Pick your bedroom. The Bronx. San Juan. Santo Domingo. Lagos. Peckham. The room has a mattress on the floor, a laptop with a cracked hinge, and a $99 interface bought used.

She makes a song. The song is good. Good in the specific way that only comes from someone who grew up inside a sound rather than studying it.

Then the machine takes over. The song gets used in a video. The video spawns a dance. The dance spawns ten thousand more. A phrase from the second verse becomes a caption, then a T-shirt, then a beverage company’s summer campaign. A fashion brand builds a whole lookbook on the aesthetic. Three AI music tools ingest the track, along with fourteen million others, and learn what that sound is made of. A streaming platform sells advertising against every second of the attention. An agency wins an award for the campaign.

Ninety million people encounter her work.

She gets a deposit. $43.17.

Here is the part that should bother you. Every entity in that chain behaved rationally. Nobody committed fraud. Nobody stole anything a court would recognize. The system worked exactly as designed, and the person who created the only thing anybody actually cared about got a number she could spend at a chain restaurant.

We were told the internet would kill the gatekeepers. What it did was replace a hundred medium-sized gatekeepers with five enormous ones that operate at the speed of light, own the data, and never take a meeting. And now we have added a new participant to the table: a machine that can consume her work, reproduce its recognizable elements, and compete for the next dollar of attention against her.

The Number That Should End the Argument

In February 2026, UNESCO published the fourth edition of Re|Shaping Policies for Creativity, a decade-long monitoring effort covering more than 120 countries and over 8,100 policy measures. It is not an activist pamphlet. It is the most boring possible document, which is exactly why you should read it.

The headline: by 2028, generative AI is projected to put 24% of music creators’ revenues and 21% of audiovisual creators’ revenues at risk. In euros, roughly €4 billion a year gone from music, €4.5 billion a year gone from audiovisual. The underlying economic study, by CISAC and PMP Strategy, puts the cumulative hit at about €22 billion over five years.

Now read that again and notice what it does not say.

It does not say people will stop listening to music. It does not say films will stop being watched, that fashion will stop mattering, that food culture will contract, that anyone anywhere will consume less culture. The opposite is happening. More music is uploaded to streaming platforms in a single day now than most labels released in a decade. Deezer reported that in June 2026, more than half of the tracks uploaded to its platform on a given day were fully AI-generated. Around 90,000 songs a day. Those tracks account for 1% to 3% of actual streams, which tells you everything about the shape of this market: infinite supply, fixed attention, and a payout pool that gets divided by a larger denominator every quarter.

So the 24% is not a demand problem. Demand for culture has never been higher in human history.

It is a distribution-of-proceeds problem.

We are not facing a shortage of culture. We are facing a collapse in the connection between cultural value and creator compensation.

Follow the Equity

In any economy, the interesting question is never who makes the thing. It is who owns the toll booth.

Line up the participants in our producer’s viral moment and ask what each one walked away holding:

This is not a morality tale. It is a balance sheet.

The platform does not need to own every song. It needs to own the audience. The AI company does not need to become the artist. It needs to own the tool that produces artist-like output at scale. The brand does not need to create the cultural movement. It needs enough permission, access, or ambiguity to commercialize it before the movement notices.

The creator makes the thing people care about. Everyone else owns the infrastructure required to convert that caring into recurring revenue.

Recurring is the word. Her income was an event. Theirs is an annuity.

Visibility Is Not Ownership: Going Viral Is Not Brand Protection

The creator economy sold a lie so effective that people repeat it in their own bios: exposure is power.

Exposure is not power. Exposure is a lead. It is the top of somebody’s funnel, and if you did not build the funnel, it is probably not yours.

Here is the honest ledger.

Going viral is an event. Ownership is an asset. A million views can vanish from public consciousness inside of nine days. A registered name, a controlled catalog, a direct customer list, a negotiated license, or an equity position keeps producing after the trend dies.

This is the difference between an audience and an asset, and it is why brand protection is a business decision rather than a legal chore. The tell is simple. Ask any creator or founder what happens to their business if the platform suspends the account tomorrow morning. If the honest answer is “it ends,” they do not have a business. They have a booking on someone else’s stage.

What AI Actually Changed

Most of the commentary treats generative AI as a copyright story. That framing is too small, and it lets everyone argue about fair use while the actual economics move underneath the table.

What changed is the cost of substitution.

Before generative AI, if a company wanted something that sounded like a particular place, looked like a particular community, or spoke in a particular cadence, it had to hire a person from that place, community, or cadence. Not out of virtue. Out of necessity. The knowledge was embodied. It came attached to a human being who had a rate.

Now the company can generate a track with the characteristics of the genre. Visuals influenced by a recognizable community aesthetic. Copy written in the linguistic register associated with a region or a diaspora. A synthetic personality that resembles a human creator closely enough for a fifteen-second ad. Two hundred variations, tested against each other, at a marginal cost of roughly nothing.

Previous technology helped creators distribute their work. Generative AI can use their work to reduce the demand for them.

Read the UNESCO finding on this and it lands harder: 79% of the creative professionals surveyed consider AI a threat to art workers. Not a tool. A threat. That is not a fringe position anymore. That is the field.

And the tell that this is an economic question rather than a philosophical one is that the input already has a price when someone is forced to pay it. In September 2025, Anthropic agreed to pay $1.5 billion to compensate book authors whose works had been taken from pirated sites for training. In November 2025, a German court found OpenAI had infringed copyright by training on songs by German musicians. The market has now demonstrated, twice, that the training data was never free. It was just unbilled.

None of this is an argument for rejecting AI. Refusing the technology is not a strategy, it is a retirement plan. The question is narrower and more useful: do creators have consent, attribution, compensation, transparency, and bargaining power, or do they have a terms-of-service update and a form letter?

One in 148

Here is the single most damning statistic in the UNESCO report, and it has nothing to do with music.

Between 2016 and 2023, 148 AI-related bills were adopted across 128 countries. Exactly one identified culture as its primary subject matter.

One.

UNESCO’s own phrase for this is that culture “remains a blind spot in AI governance,” and the report goes further, noting that the Global Digital Compact overlooks the diversity of cultural expressions entirely, and that the OECD AI Principles contain no mention of cultural expression at all. The omission is striking for a specific reason: cultural content is the primary training material for these systems. We wrote comprehensive rules for the engine and forgot to mention the fuel.

Why does this happen? Because policymakers file culture under entertainment. Nice to have. Ministry of Fun. Something you fund when the budget is healthy, which explains why public funding for culture globally sits below 0.6% of GDP and has been declining.

That filing error is expensive, because culture is not entertainment. Culture is economic infrastructure. It determines language, purchasing behavior, tourism flows, what fashion sells, what food gets ordered, what music soundtracks a campaign, what a political identity feels like, and what any given product is perceived to be worth. Culture is why one hoodie costs $40 and a materially identical hoodie costs $400.

Culture creates markets. Treat its raw material as a free input and you have written a policy that protects the technology’s economic upside while quietly externalizing the cost onto the people who generated the value.

Policymakers regulate AI as technology. Creators experience it as a labor market, a licensing market, and an ownership system.

Where Brand Value Leaks: Five Failures I See as a Trademark Attorney

I have spent my career on the paperwork side of this, filing and defending federal trademark registration for apparel brands, beauty founders, food and beverage operators, and creative agencies. The pattern is boringly consistent. Brand value rarely leaks because the work was not good enough. It leaks through five specific holes, and every one of them is a brand protection failure rather than a creative one.

There is a sixth leak that operates at the level of entire countries. UNESCO found that while collective management organizations for music exist in about two-thirds of member states, only 23% to 24% of developing countries have collecting societies covering audiovisual, literary, visual arts, photography, or design. If there is no infrastructure to collect the money, the money does not go uncollected. It goes to someone with better infrastructure.

Value does not leak because the work lacks demand. It leaks because someone else controls the name, the contract, the platform, the data, or the pipe.

The Ownership Stack: How to Protect a Business Name Before It Scales

If you build anything cultural, whether that is a catalog, a clothing line, a skincare brand, a restaurant group, or a creative studio, you are building on five layers. Most people build one and hope.

The strongest creator businesses do not merely produce content. They build a stack of assets around the attention the content generates. Content is the customer acquisition cost. The stack is the company.

The Part Nobody Wants to Say Out Loud

Now the honest complication, because a column that ends with “register your trademark” is a brochure, not an argument.

Not everything valuable can be individually owned, and it should not be.

Culture is cumulative. It is built from borrowed rhythms, inherited language, communal aesthetics, religious forms, regional slang, and stories that predate every person currently profiting from them. Nobody invented the clave. Nobody owns a color palette that came out of a neighborhood. If we resolved this problem by converting every cultural influence into private property, we would end up with a licensing regime that strangles the very exchange that produces culture in the first place. Genres would become estates. Traditions would become litigation.

But the opposite conclusion is worse, and it is the one currently winning by default: that cultural material is free for the most capitalized company in the room to extract, scale, and monetize with no obligation whatsoever to its source.

So the real question is harder than either camp admits. How do we preserve cultural exchange without turning communities into unpaid suppliers of raw material?

That is where individual ownership and collective stewardship have to do different jobs. An individual creator can and should own their name, their catalog, their contracts, their audience. A community cannot own a rhythm, but it can insist on infrastructure: collecting societies, transparency requirements, labeling of synthetic content, consent and compensation standards for training data, and cultural provisions inside AI law rather than cultural provisions as an afterthought in one bill out of 148.

One of those is a business decision you make on a Tuesday. The other is a policy fight that takes a decade. Both are necessary. Only one of them is under your control this week.

Own More Than the Content: When to Trademark Your Brand

Back to the bedroom.

The song still went viral. The dance still happened. The phrase is still on shirts. The platforms still sold the ads, the brand still got its campaign, the model still trained on the track, and none of that is getting reversed.

The only variable that was ever in play is what she owned when the attention moved on.

That is not a moral failure on her part. Nobody hands out this playbook, and the people who understand it best are generally the ones on the other side of the table. But it is learnable, and it is cheap relative to what it protects.

Culture is not becoming less valuable. It is becoming easier to extract.

The next generation of creators will not win by producing more. Volume is the one thing the machines have already beaten us at, permanently, and it is not close. They will win by owning the names, the rights, the relationships, and the systems that turn cultural attention into durable economic value.

Before You Go: The Questions Founders Actually Ask Me

Short answers to the things that come up in almost every consultation. None of this is legal advice for your specific situation.

If you are building in apparel, beauty, food and beverage, or creative services and you want to know where your own leaks are, book a consultation.

Sources