Mexico vs. Technology: One Law at a Time

View on LinkedIn

At Mexico City's International Airport there is a system that no one designed on purpose, yet it operates with an almost admirable precision. You arrive at the terminal, look for transportation, and discover that your options have been carefully limited. The authorized taxis charge fares that are double or triple the market rate. Platforms like Uber operate in a legal gray zone — harassed, blocked from access roads, hounded by organized groups of drivers who defend their territory with the same ferocity a medieval guild used to protect its trade routes.

The taxi drivers' argument is always the same: we were here first. We invested. We have rights. And the argument is true — right up until you ask: rights over what, exactly? Over the passenger?

I raise this because in Mexico a legal regime is being built, piece by piece, that does to artificial intelligence exactly what the airport taxi drivers do to transportation: it locks the doors and charges more. And the first sector being used as the testing ground is dubbing.

In April 2025, a legislator introduced a bill amending the Federal Copyright Law, the Federal Film Law, and the Federal Labor Law. In February 2026, the Executive announced an even broader package. The stated subject: protecting dubbing actors against artificial intelligence.

The underlying problem is real. A voice actor's voice is part of their identity. Cloning it without consent is a violation of their rights. No reasonable person disputes that. And the legal solution is straightforward: a voice should follow the same logic as the right of publicity — a personal attribute whose reproduction requires authorization from the holder. That can be resolved with a surgical amendment to the law. Two paragraphs. One article.

But that is not what these bills propose.

What they propose is a regulatory apparatus that does not merely protect the voice of the person to whom it belongs, but instead erects a wall of restrictions around anyone who wants to use artificial intelligence to produce voice — even if they are not cloning anyone, even if they are not imitating anyone, even if they are using a legitimate tool for a legitimate purpose.

Let us consider a concrete example. An educational content creator who wants to narrate courses in six languages. An independent studio in Coyoacán producing documentaries. A startup developing accessible audiobooks for visually impaired users. None of them are stealing any actor's voice. They are using technology.

Here is what they would have to do if these reforms pass:

Six barriers to using a tool. Not to clone someone's voice — that should be illegal, full stop — but to use artificial intelligence in voice production.

The labeling requirement deserves a moment's pause, because it is where the true intent is revealed.

There is something psychologists call the "identifiable victim effect." When you can put a face to someone who is suffering, the emotional response is disproportionately strong. A dubbing actor who loses work because a machine cloned their voice is a powerful story. And behind that story, the legislator slips in an obligation that has nothing to do with protecting the actor: labeling.

Think about it. We do not label books written with word processors. We do not label music produced with Auto-Tune. We do not label films that use CGI to recreate entire cities, armies, or planets. We do not require Pixar to post a notice saying "no real actor participated in this film." But if you use AI for voice, you carry a mark.

Mandatory labeling is not transparency. It is a regulatory stigma designed to make AI-generated content appear inferior, suspect, and second-rate. It signals to the market: distrust this. It is the exact equivalent of requiring Uber to display a notice saying "this is not an authorized taxi." It does not inform the passenger; it protects the taxi driver.

And beyond that, it is technically absurd. The bill demands disclosure of whether voice was "assisted or performed" by AI. The difference between those two words is enormous. A voice actor who uses software to clean up background noise is being "assisted" by AI. A system that generates a fully synthetic voice is "performing" the narration. The law treats them as if they were the same thing. It does not define the boundary. It does not explain who verifies compliance. It does not say what happens when a small studio cannot afford the compliance costs this entails.

The question nobody asks: If a musician uses AI to compose tomorrow, do we label the album? If an architect uses AI to design, do we label the building? If a lawyer uses AI to draft a contract, do we label the contract? The labeling logic is unsustainable the moment you apply it to any other sector.

The full inventory of the apparatus makes the pattern impossible to ignore.

In 1997, economist Mancur Olson published an observation that changed the way we think about regulation. Olson noted that small, organized groups almost always defeat large, dispersed groups in political battles. The reason is mathematical: if you are one of five hundred taxi drivers who benefit from a barrier, each driver gains a great deal. If you are one of thirty million users who pay more because of that barrier, each user loses very little. The taxi driver has every incentive to organize and block streets. The user has every incentive to complain on social media and get on with their life.

Olson called this "the logic of collective action." Mexico turned it into a form of government.

What makes the dubbing package worrying is not only what it does to the audiovisual sector. It is the precedent. Because if it works — if the dubbing guild secures a regulatory wall with six barriers and zero definitions — every other guild is going to want one of its own. And we are already seeing it: the same 2026 legislative package includes mandatory national film exhibition quotas, government reviews of theatrical programming, and visibility obligations for digital platforms. The architecture is identical: protect the supplier, not the user.

And there is a fact that should concern the drafters of these bills.

In General Declaration of Unconstitutionality 2/2024, the Supreme Court of Justice of the Nation examined Article 8 of the Federal Film Law — the article governing original-language versions, dubbing, and subtitling — and found that it imposed disproportionate restrictions on freedom of commerce. The goals were legitimate, but the means were excessive. It did not survive the proportionality test.

Any regulation imposing barriers to the use of technology in audiovisual production will face the same standard. And a law that prohibits broadcasting AI-generated content without meeting requirements the law itself leaves undefined, that compels labeling without any technical criterion, that creates "ethical use" obligations without defining what "ethical" means — that law has a serious constitutional problem.

There is a thought experiment I find indispensable. Let us take the logic of these bills and project it five years forward.

When AI threatens software developers' jobs — and it is already doing so — will we enact a provision prohibiting the generation of code without a programmer's consent? Will we require all "AI-assisted" software to be labeled? When graphic designers are displaced by generative tools, will we mandate marks on every image? When autonomous vehicles arrive on Mexican streets, will we protect Uber drivers with the same barriers that taxi drivers used against them?

If Mexico's response to every technological disruption is to build a regulatory wall around the affected guild — consent requirements, labels, ethical audits, royalties without a defined beneficiary — we do not have an innovation policy. We have a preservation policy. And what is being preserved is not jobs, nor culture, nor the dignity of work: it is the existing power structure.

I return to the airport.

What makes the taxi system work is not that taxi drivers are bad people or that their service is terrible — although it frequently is. What makes it work is that the regulation is designed to protect the supplier, not the user. The passenger has no voice in how the system is designed. Their interest — getting there quickly, safely, and at a fair price — never appears in the regulatory equation.

The AI and voice bills replicate that architecture exactly. The dubbing guild's interest appears in every paragraph, every article, every new sub-provision. The interest of those who want to use technology legitimately — the content creator, the independent studio, the educational startup, the audiobook producer — appears nowhere.

Protecting voice as a personal attribute is correct and necessary. It can be accomplished with a minimal, proportionate, and surgical reform, just as the right of publicity was. What is not correct is using that protection as a pretext to build a regime of barriers, labels, and permits that regulates the tool rather than the abuse, that stigmatizes the technology rather than sanctioning the violation, and that responds to disruption with the same strategy as the airport taxi drivers: lock the doors and charge more.

Mexico does not need more laws written for guilds. It needs laws written for citizens.