Just over a decade ago, superhero movies were at their absolute peak — a cultural phenomenon that completely redefined what we expected from a night at the theater. I remember watching “Avengers: Age of Ultron,” where the ultimate villain was Ultron: a rogue artificial intelligence that gained self-awareness, built itself a physical robotic body, and posed an existential threat to humanity and the world as we knew it.

Today, barely 10 years later, reality has taken a fascinating turn. AI is no longer that sci-fi antagonist threatening us from the screen. Instead, it has slipped quietly into our daily lives: writing scripts, generating visual characters, and mimicking voice actors to adjust accents, languages, and cadence with a precision that comes dangerously close to what we naturally produce.

This brings us to a compelling shift in perspective. We are no longer dealing with a distant cinematic threat, but with an everyday reality that is changing — for better and for worse — how we understand the world and how we produce creative works.

Naturally, this has become one of the most frequent conversations I have with my peers in the legal sector. We constantly debate ethics, creativity, and where the human element fits when creating new work alongside AI. Everyone in my field seems to be having some version of the same fundamental question right now: can a company actually own what a machine helped make?

The honest answer is usually: it depends.

The Engine Behind Companies

Artificial intelligence stopped being a futuristic idea somewhere in the last two or three years and turned into the engine behind how companies write, design, price, and predict what their customers will do next. The speed is genuinely impressive. What concerns me, from where I sit as a lawyer working in regulatory matters, advertising, e-commerce, and data privacy, is a habit I keep seeing: treating the tool’s output as if it were automatically true, automatically compliant, and automatically good enough.

It isn’t. And the gap between what a model can produce and what a company can legally claim or safely publish is exactly where my work happens.

Ethics applied to artificial intelligence is not a slogan for a sustainability report. It is governance, it is reputational risk management, and it is a responsibility that no executive can quietly hand off to a chatbot.

Regulators around the world are not moving at the same pace, which makes this harder than it should be. The European Union has its AI Act. The OECD’s broader AI Principles, adopted by more than 40 countries, sit alongside a separate, narrower set of due diligence guidelines the organization issued this past February. UNESCO’s 2021 Recommendation on the Ethics of Artificial Intelligence, the first international instrument on the subject, added equity, non-discrimination, and human oversight to that mix. In the United States, a federal posture that favors technical development coexists uneasily with state laws in places like in California and Colorado.

The Mexican Landscape

Mexico, for its part, doesn’t yet have a single AI statute — there are proposals in Congress with a risk-based approach, and even talks of amending Article 73 of the Constitution — but that doesn’t mean the space is unregulated. Companies here are already bound by a broader legal framework that includes the Constitution itself, the Federal Law for the Protection of Personal Data, the Federal Copyright Law, the Federal Law for the Protection of Industrial Property, the Federal Consumer Protection Law, and, in sectors like the ones I work in, the General Health Law, among others.

The absence of an AI-specific law is not the same as the absence of law.

On top of that statutory floor sit self-regulatory frameworks that, in practice, function as the day-to-day compass: the Chapultepec Principles issued by the federal government’s science and digital transformation authorities, and the International Chamber of Commerce’s Code of Advertising and Marketing Communications. Both boil down to the same demand — transparency, honesty, decency, and respect for the audience, regardless of whether a human or an algorithm produced the content.

The blind spot I run into most often involves who actually owns what an AI tool helps create. There’s a comfortable assumption inside a lot of companies that anything a platform generates automatically becomes company property, the same way an employee’s work product would. Legally, that assumption doesn’t hold.

Courts in multiple jurisdictions have started to address this issue. For example, Mexico’s Supreme Court already in Direct Amparo 6/2025, held that artificial intelligence cannot be an author — that status belongs exclusively to natural persons — which means a work produced in a fully automated way carries no copyright protection at all as a general rule. Germany reached a similar conclusion from a different angle: the Munich District Court (AG München, 142 C 9786/25) found that a generic, open-ended prompt, without identifiable human creative input layered on top of it, isn’t enough to support a claim of intellectual ownership. Two courts, two legal systems, the same underlying instinct. Worth being precise here, too: none of this means an AI-generated work automatically falls into the public domain. It just means nobody can claim rights on it, which, from a competitive standpoint, may be worse.

Creating a Habit

For agencies and brands, that instinct has to translate into a habit, not just a legal footnote. If a team asks a model to produce a campaign or a logo without reflecting enough human creative influence, the author’s personality through free and creative choices and nobody logs what happened throughout the process — the iterations, the edits a human actually made, which version got chosen and why — the company has no protectable claim if a competitor copies the result tomorrow. Keeping a simple record of that creative process, a kind of running log, is currently the only realistic way to document the human authorship that intellectual property authorities will ask about later.

There’s a second problem that worries me just as much, and it’s harder to fix: bias. These systems learn from historical data, and historical data carries the prejudices of the moment it was collected in — about gender, race, socioeconomic status — and it will reproduce them unless someone actively audits for it beforehand. Studies looking at advertising in Mexico keep finding the same patterns: stereotyped roles, a version of the population that looks nothing like the actual one. Plug a language or image model straight into a campaign without checking its output first, and you’re not avoiding those distortions — you’re mass-producing them.

Then there’s the matter of hallucinations, which I think is the most dangerous because of how convincing it sounds. These models are probabilistic; they don’t know things, they predict what a plausible answer looks like, and they say it with total confidence whether it’s accurate or invented. A citation to a legal precedent that doesn’t exist, a statistic pulled from nowhere, a historical fact quietly altered — any of these can put a company in real legal and reputational trouble if nobody bothers to check.

‘Tone at the Top’

Which is why I keep coming back to something people in compliance call “tone at the top.” The rules for how a company uses these tools, the protocols that keep personal data, confidential information and trade secrets out of a vendor’s training set, the contractual language that stops a supplier from feeding client’s confidential information into a public model — none of that gets decided by the software. It gets decided, or it doesn’t, by the executives running the company.

A tool being technically capable of something has never been the same question as whether a company should do it. Before any of this gets implemented, leadership needs to ask whether it complies with the law as it stands, whether it respects people’s privacy, and whether it’s honest with the customer on the other end. What a model can generate still has to pass through judgment, decency, and a sense of responsibility that simply can’t be delegated.

Artificial intelligence is a genuine asset in daily work. We no longer see it as a fictional threat, but as a tool whose potential we must continue to leverage—even within an ethical and regulatory framework that is still developing. Expert recommendations are valuable, but above all, we need the common sense that keeps us human. That capacity to discern, create, and truly convey what we have inside into our creative work—whether it’s a fictional piece, a narrative, a corporate logo, a presentation, or a way of seeing the world—is what continues to separate us from the tools we use.