AI in Mexico: Innovator or Imitator in the Global Scenario?

Introduction: Mexico without its own model

Mexico is not late to artificial intelligence regulation; it lands at an uncomfortable crossroads: when copying is easier than thinking. While the European Union created the first major legal framework for AI based on risks and fundamental rights, the United States bets on innovation free from excessive regulations, and China exerts firm control over information and national security, the Mexican vision is still undefined. This context not only reveals a lag but a historic opportunity to propose a bold path and create an artificial intelligence public policy: Mexico can be a proactive actor in AI governance, boosting its economic and social development, instead of merely emulating external recipes through the creation of an AI regulatory framework.

On April 3, 2026, the Official Gazette of the Federation published a reform to the Federal Law for the Protection of Industrial Property (LFPPI). More than two hundred articles modified. But a conceptually profound change fits into two silent movements that any hurried reading misses: the expansion of the law's very object, in Article 2, and the redefinition of the powers of the Mexican Institute of Industrial Property (IMPI), in Article 5.

In Article 2, the legislator incorporated for the first time as a legal purpose the "promotion and encouragement of technology transfer." In Article 5, it turned IMPI into a mandatory legal advisor on licenses, assignments, and transmissions of rights; imposed cooperation with the newly created Secretariat of Science, Humanities, Technology, and Innovation; and mandated the promotion of industrial property regulatory compliance systems among productive sectors through training, awareness, and education programs. IMPI ceased to be a mere rights registrar to become, at least on paper, an active agent of innovation.

However, for thirty-three years, the institute responsible for administering the industrial property system of one of the planet's ten largest markets did not have, as an explicit legal function, technology transfer, articulation with the scientific authority, or the obligation to build a compliance culture within the productive fabric. South Korea institutionalized that package of functions in the 1960s. Israel, in the 1980s. China, in the 1990s. Mexico discovered them, all at once, in 2026.

This symptom describes the deepest problem of Mexican public policy regarding artificial intelligence and, in general, any branch: we do not legislate late because we lack jurists. We legislate late because, when it comes to technology, the Mexican State came to understand late exactly what it must do.

Global AI regulation and its rationale

Global AI regulation is advancing forcefully. The EU enacted the AI Act, the world's first attempt at a "comprehensive legal framework for AI." This regime classifies AI systems into risk levels (from unacceptable uses to minimal risk) and explicitly prohibits the most harmful ones, such as subliminal manipulation or mass surveillance. Designed to "foster trustworthy AI," the European Act ensures safety and fundamental rights while boosting innovation through a broader package of measures (funding, AI Factories, etc.).

Meanwhile, the United States does not yet have a specific federal law, but its executive agenda repeats the motto "nothing that stops American supremacy in AI." A recent decree (EO 14179, January 2025) orders the review and revocation of policies that limit innovation, declaring that national policy must be "minimally burdensome" to maintain technological leadership. The goal is a single standard across the country (avoiding the "maze" of 50 state regulations) that protects competition and does not politicize it.

Given the federal vacuum, states took the initiative: Colorado passed its AI Act in 2024—though it later postponed its implementation—California enacted the Transparency in Frontier Artificial Intelligence Act in September 2025, and Texas signed TRAIGA in June of that year. On December 11, 2025, Trump responded with a new executive order creating an AI Litigation Task Force in the Department of Justice with an explicit mission: to sue states that legislate on AI. Days later, New York passed the RAISE Act. Innovate first, regulate perhaps never, and if states regulate, challenge them.

At the extreme, China combines a State strategy: after a frenzied push phase until 2022, it has intensified regulation to control the flow of information and reinforce ideological security. During the "Crackdown Era" (2020-2022), the government imposed rules focused on monitoring recommendation algorithms and censoring content, prioritizing political order over Western ethical considerations. Only recently, after losing pace to ChatGPT and its economy slowing, has Beijing slightly relaxed its iron grip; however, its objective remains technological sovereignty and social stability.

At the international level, organizations such as the OECD and UNESCO have outlined shared principles: transparency, accountability, and respect for human rights stand as fundamental pillars. The OECD AI Principles promote "innovative and trustworthy" AI that respects democratic values, focusing on inclusive growth, safety, and accountability. UNESCO, for its part, insists that the protection of human dignity and rights is "the pillar" of the global ethical framework, translating those values into policies in education, gender, environment, etc. In short, although each region highlights its own emphasis (risk in Europe, market in the US, control in China), there is a tacit consensus on human rights, equity, and accountability as common bases.

Mexico: Between historical imitation and strategic opportunity

Unlike pioneering powers, Mexico has barely charted a course. Since 2020, there has been a wave of federal legislative initiatives, but none have prospered. In the previous legislature (2021–2024), laws were presented with various approaches: an ethico-robotic law (April 2023) focused on human rights and an ethics council; proposals for constitutional reform to empower Congress over AI and cybersecurity; a national AI agency (October 2023) inspired by European experts; and risk-based rules in the style of the EU (projects from November 2023 and April 2024). For example, the April 2024 project established three risk levels (unacceptable, high, and low) and mandated preventing the spread of "harmful content" (disinformation, hate speech), in addition to requiring deepfakes to be marked. Another initiative (October 2024) repeats the EU risk model and adds guiding principles—ethics, legality, transparency, and use without ideological bias—as well as creating a National AI Center to coordinate public policy.

However, these proposals remain scattered: several have been archived without discussion. Beyond the number of initiatives, there is a prevailing absence of a coherent framework. Many literally copy European categories or abstract lists of "principles" (such as autonomy, justice, non-discrimination) without addressing how to implement them. For example, the new law proposed in February 2026 lists principles—from confidentiality to multi-sector innovation—all "aligned with international standards," but implicitly recognizes that without their own structure, they would be mere empty statements. The lack of strong institutions illustrates this: there is currently neither a national AI plan nor a unified governing body.

None of these pieces are yet connected to the others. There has been no National Artificial Intelligence Strategy in effect since 2018, when the Peña Nieto administration presented—with British advice—the tenth such document in the world and the first in Latin America. That document was de facto abandoned during the following six-year term. The IA2030Mx coalition, from civil society, tried to keep the effort alive but did not succeed.

Superimposed on this scene is a hard fact: in March 2025, seven autonomous constitutional bodies disappeared, including INAI (the data protection authority). Their functions were absorbed by a Secretariat of Anticorruption and Good Government that has, at best, a one-year learning curve to manage the data protection governance of a country discussing how to regulate AI.

The discussion on artificial intelligence in Mexico is not limited to the lack of a specific law: it also appears in concrete decisions on personal data, technology providers, and legal responsibility, as developed in the analysis of AI privacy filters.

In all of this, there is a contradiction that cannot be hidden: a National AI Laboratory is announced in the same six-year term in which the data protection oversight body is dismantled; a supercomputer is built while spending on research and development remains around 0.4% of GDP, the lowest in the OECD; legislative initiatives that look very much like the European AI Act are discussed while ignoring that Mexico—having participated as an observer—has not signed or ratified the only binding international treaty that exists on the matter.

It is not that Mexico is doing nothing. It is that it is doing many things, all halfway, without a common thread.

In contrast, we find ourselves at a crossroads: will we be mere imitators of foreign norms, or will we take the opportunity to innovate institutionally?

The problem is not a lack of law, but a lack of vision

The comfortable diagnosis says that Mexico lacks an artificial intelligence law. It is true. But it is a superficial truth, like saying a patient lacks a medicine without first asking if they have a diagnosis.

Let's return to the new design of the LFPPI. Technology transfer, cooperation between the industrial property office and the scientific authority, and the construction of a compliance culture through training and education for the productive sector are not new ideas. South Korea institutionalized them with the Korea Institute of Science and Technology in 1966, when its per capita income was lower than Mexico's. Israel articulated them with the Yozma Program in 1993: one hundred million initial dollars that, in a decade, unleashed a venture capital industry of three billion two hundred million and, above all, the transformation of a small, resource-poor country into the world's leading nation by per capita unicorn valuation. Singapore did it with A*STAR. Estonia, with X-Road and digital identity since 2001. Taiwan, with ITRI since the 1970s.

Mexico had, at the time, the same intuition. The Mexican Petroleum Institute, founded in August 1965 at the initiative of Jesús Reyes Heroles as Pemex's technological arm, developed its own patents in hydrodesulfurization and catalyst regeneration and was for decades the largest Mexican applicant before the IMPI. Cinvestav, founded in 1961 by Arturo Rosenblueth—a collaborator of Norbert Wiener and co-author of the foundations of cybernetics—was the bet on creating our own science. Guillermo González Camarena had patented, in 1940, the Sequential Field Trichromatic System: one of the world's first color television systems. Luis Miramontes co-invented the compound that gave rise to the contraceptive pill in 1951 at Syntex-Mexico.

That generation understood something that the subsequent political class forgot: that innovation is not regulated; it is built. And building it requires permanent institutions, not announcements every six years.

Between 1954 and 1970, during the so-called Mexican Miracle, the Mexican economy grew at 6.56% annually. Between 1940 and 1970, the State provided scholarships to some thirty thousand Mexican students abroad. IPN, created in 1936, continues to lead national academic patenting today. All of that was dismantled with a series of accumulated decisions: the 1982 crisis, trade liberalization without a technological transition policy, the privatizations of the 1990s that produced private monopolies instead of competition, and the sustained failure to fulfill—and subsequent elimination of—the legal mandate to allocate 1% of GDP to science and technology.

The numbers we inherit demonstrate the obvious. Mexico ranks 56th out of 133 in the Global Innovation Index 2024. Third in Latin America, behind Brazil and Chile. More efficient in innovation outputs—position 52—than in inputs—position 73—which is a diplomatic way of saying we do more with less because we have less. Structural weaknesses are in three areas: institutions (rank 106), infrastructure (rank 71), and human capital (rank 63). There is not a single Mexican scientific-technological cluster among the world's main ones.

Meanwhile, on the other side of the border lives one of the most qualified populations of Mexican professionals on the planet. INEGI estimates that there are around one and a half million Mexicans with higher education outside the country, more than 80% in the United States. The International Organization for Migration raises the number of highly qualified Mexicans abroad to over 1.3 million. Approximately one in four Mexicans with a doctorate lives abroad. 85% of postdoctoral stays in exact sciences are carried out abroad. There are Mexicans building AI models in Mountain View, Seattle, London, and Zurich. We are not calling on them.

In 2024, IMPI granted 10,897 patents. Of these, 694—6.4%—went to Mexican residents. A record for the last thirty years—a record that was broken again in 2025. To put it in perspective: in 2023, China received around 1.64 million invention patent applications; the United States, more than 500,000; Japan, more than 400,000; South Korea, nearly 290,000; Mexico, just over 30,000.

A historic opportunity that cannot be wasted

In every perceived risk, there is a hidden opportunity. Rather than focusing the debate on fear of AI, Mexico could propose a proactive vision. Instead of thinking of AI only as a danger (which, of course, requires safeguards), we must imagine it as the backbone of innovative ecosystems. For example, the country could support the creation of regional technological clusters with an AI focus, linked with key industrial sectors (health, agribusiness, energy, justice, etc.). These clusters would link universities, companies, and the government in living laboratory networks, in the style of European AI Factories.

It is also advisable to promote open access to data and capabilities: Mexico can incentivize public repositories of anonymous information (big data on health or mobility) and offer tax credits for infrastructure (clouds or supercomputing). Thus, local SMEs and startups can experiment instead of depending on foreign algorithms. In addition, AI must be integrated into education: from basic digital literacy to the training of data scientists and digital ethics at the graduate level. Incorporating AI into study plans will ensure future technical cadres, avoiding the brain drain we already suffer in engineering.

In the public sphere, the State could offer innovative applications as a proof of concept: for example, AI-based judicial assistants to speed up trials or predictive systems in public safety, always with rights assessment. These initiatives would demonstrate the social utility of AI and build trust. Finally, regulation should be aligned with the digital economy: updating intellectual property laws for AI, promoting data interoperability agreements (data trusts), and ensuring competition in the AI market (to prevent Big Tech monopolies), as competition experts already suggest.

In short, public policy must shift toward an enabling approach. It is not an impossible task: INDAUTOR, IFT, or other agencies could quickly adapt to oversee AI (to date they regulate digital markets and intellectual property, respectively). Together with an independent National AI Council (as proposed in recent legislation), a sanhedrin of innovation would be formed to prescribe regulations without losing sight of economic growth. The goal would be clear: turn the current lag into a competitive advantage, using AI to solve local problems (poverty, injustice, government efficiency), not just to fulfill foreign regulatory checklists.