Mexico: 200 AI Bills, No Federal Law

Operation Timeline

Attack Chain and TTPs

The material does not describe an intrusion campaign, a single malicious operation or a discrete technical incident. There are no published IOCs, no hashes, no domains, no IP addresses, no attributed malware and no concrete exploitation vectors. For that reason, a classic attack-chain reconstruction would not be honest. What the research does allow is a different reading, one focused on a chain of regulatory and operational exposure that leaves companies, states and citizens with uneven coverage against AI and cybersecurity risks.

That chain starts with the lack of a general framework. In Mexico, the absence of a General Artificial Intelligence Law leaves each state to produce its own response, often in the form of criminal-law initiatives or incomplete definitions. At the federal level, the lack of a general corporate cybersecurity law means obligations are concentrated in the state and then cascade to contractors, suppliers and sectors regulated by other rules. The result is a compliance surface that is hard to map, with different criteria across jurisdictions and gray areas in the value chain.

In that context, the relevant TTPs are not malware-based but policy- and compliance-based. The most visible risk technique is regulatory fragmentation. Yucatán criminalizes AI uses that generate or manipulate fake content, while San Luis Potosí is moving to unwind sanctions. The state-by-state count shows that only two states explicitly defined what they mean by AI. That means the same system can be treated very differently depending on the jurisdiction. For a company operating in several states, the problem is not only legal. It is also operational, because internal policies on AI use, evidence retention, traceability and incident response can be subject to uneven local rules.

The second exposure layer is opacity around in-house and third-party systems. El Economista says that, without a general law, companies operate under sector-specific obligations with limited oversight over the AI systems they use. That points to a concrete weakness. Many organizations consume third-party tools, models embedded in cloud services, or automations integrated into enterprise software without a robust inventory of models, data, purposes and risks. Without a general standard, the definition of minimum controls gets scattered across privacy, labor, copyright, consumer issues and supplier contracts.

The third layer is the asymmetry between innovation and supervision. ITSitio and Expansión point to a federal architecture that proposes a National AI Council, an algorithmic audit platform and a risk traffic-light system. That suggests an attempt to move away from a purely reactive model toward one based on classification and technical oversight. But the material does not yet show the legal translation of that proposal. It also does not show a consolidated operating authority, audit procedures or published rules for interoperability with states, companies and sectors.

Normative TTP Matrix

Regional Impact

Regional Overview

The regional comparison shows that Mexico’s problem is not a total lack of activity, but the absence of a regulatory closeout. Peru already deployed a specific AI law and a regulation that took effect in January 2026, with concrete obligations for high-risk systems and explicit prohibitions on mass surveillance without legal basis, individual crime prediction and autonomous lethal systems. Chile is processing a risk-based AI bill and pairs it with Personal Data Law 21.719, which governs automated decisions and profiling. Uruguay combined reforms on state cybersecurity, a national AI center and specific rules for automated systems in delivery and transport platforms. Mexico, by contrast, is still debating the core framework while continuing to accumulate state and sector-specific initiatives.

The cybersecurity contrast is similar. Uruguay requires public bodies to file annual reports on their cybersecurity status and send them to Agesic, in addition to complying with the Uruguayan Cybersecurity Framework. Chile, based on the available material, appears as a more integrated jurisdiction across data, cybersecurity and incident response. Mexico, meanwhile, is building a cybersecurity policy for its federal public administration, with future specific legislation still under development. That leaves companies in an intermediate position, with indirect obligations if they supply the state, but without a comparable general framework for all sectors.

Mexico

Mexico concentrates the largest volume of material and also the greatest regulatory dispersion. The figure of nearly 200 state initiatives, with 39 approved, reflects intense political activity but also a lack of convergence. Some states are advancing through criminal law, others are trying out definitions, others are repealing sanctions, and several have no recorded proposals at all. Yucatán chose to punish the generation or manipulation of fake images, videos or audio through AI with prison and fines of up to 234,000 pesos. San Luis Potosí moved toward repealing criminal sanctions linked to the technology. The result is a map with diverging rules for very similar uses.

At the federal level, the legislative sequence is not trivial. On July 24, the bill for the Federal Law for the Ethical, Sovereign and Inclusive Development of AI was introduced. The same text seeks a National Artificial Intelligence Council, a National Algorithmic Audit Platform and a risk traffic-light system. Days later, Gabriela Jiménez Godoy introduced a constitutional reform to Article 73 to give Congress explicit authority in the area. But the operational fact remains the same, there is no general law approved or published. In addition, the specialized publication ecosystem, including Academia de IA, Ecosistema Startup and ITSitio México, agrees that the debate is still in process.

Federal cybersecurity is moving at a different speed. The National Cybersecurity Plan 2025-2030 and the General Cybersecurity Policy for the Federal Public Administration create a clearer institutional setup for the state. There is an annual roadmap, an operations center, an incident response center, a critical infrastructure inventory and a provision for future legislation. However, the material also makes a gap visible. There is no general corporate cybersecurity law that unifies requirements for companies outside the state perimeter, nor a general AI framework that orders audits, risk classification and responsibilities on a national scale.

Peru

Peru is the clearest benchmark in the material for comparing regulatory speed. Law No. 31814 was enacted in July 2023 and strengthened with a regulation approved in September 2025 and effective since January 2026. According to the sources, the rule classifies systems by risk level, prohibits certain high-impact practices and requires human oversight for high-risk uses. In health, education, justice, security, the economy and finance, enhanced obligations begin to apply on September 10, 2026. There is also a staggered timetable for other sectors and for certain public actors.

The combination of law, regulation and deadlines makes Peru look like a more closed regulatory governance model than Mexico. The material also shows limits, because Infobae Perú reported that there is no legal authorization to use AI in administrative sanctioning. In other words, the country is not opening the door to automating sanctions without an explicit legal basis. AI can improve public services and streamline internal processes, but it cannot simply replace the legal framework in administrative punishment.

Chile

Chile still does not have an approved specific AI law, but it does have a bill in process with a risk-based approach. Ecija describes four categories, unacceptable risk, high risk, limited risk and no evident risk. Nodo.build adds that Personal Data Law 21.719, which it says has been in force since December 2024, regulates automated profiling, algorithmic decisions and recognizes a right to object to decisions based on AI. The country shows a more coherent structure between data and future AI controls, although the material also includes claims from sources that are not fully confirmed about Chile’s regional leadership in cybersecurity.

For comparative analysis, Chile works as an intermediate model, but one more structured than Mexico. It has a risk-language AI bill, a personal data law that addresses automated decisions and an incident response infrastructure recognized by LACNIC. That does not mean regulation is complete, but it does mean the baseline looks more integrated.

Paraguay

Paraguay combines regulatory lag with operational activity in cybersecurity. It does not have a promulgated AI law or an approved national strategy, although MITIC and CONACYT are developing the topic. MuchoTexto says there is no approved document, no budget allocation and no designated AI authority. At the same time, the country is seeing real incident pressure, with 8,309 reports and 4,109 cybersecurity incidents handled by CERT-PY between January 2023 and July 2026. So far in 2026, nearly half of attacks affected the private sector and more than a third hit the public sector.

The material also shows sector-specific responses. Paraguay has an electronic trust services law, a personal data protection law aligned with GDPR, and debates over e-commerce reform that seek to incorporate AI, strengthen cybersecurity and eliminate gray areas in sales over WhatsApp. Public policy exists, but it still has not turned into an integrated AI framework. That is why the country appears behind in the ILIA index and at the same time active in partial measures.

Uruguay

Uruguay shows a more mature institutional cybersecurity structure and an AI agenda focused on development. Decree 66/025 requires covered entities to comply with the Uruguayan Cybersecurity Framework, while the 2026 accountability process requires annual reports on each public agency’s cybersecurity status. On the AI side, the government launched the National Artificial Intelligence Center, with public-academic leadership and IDB funding. The country also maintains specific rules for automated systems in delivery and transport platforms, with information duties, algorithmic explanations and risk assessments.

Uruguay does not appear in the material as having a closed general AI law, but it does stand out as a country where the state has already institutionalized important parts of the issue, especially in the relationship between algorithms, labor, information security and national research capacity. That combination places it, in the material, ahead of Mexico in cybersecurity organization and closer to a structured public policy on automation.

United States

The United States still does not have a federal AI law equivalent to the European one. RecMedia describes an active debate between the federal government and the states, while Talla Política emphasizes a sectoral and competitive approach, supported by executive orders and by existing agencies such as antitrust, commerce and civil rights. In the context of the US-Mexico-Canada Agreement, Buzos characterizes the U.S. position as zero regulation, although that source stays in the opinion space of the material. For Mexico, the point matters because of trade weight and because it helps explain pressure to avoid overly rigid frameworks.

Countries without consolidated coverage in the material

There are no additional verifiable facts in the research for Argentina, Bolivia or Colombia.

Technical Indicators

No technical IOCs were published in the compiled material. There are no hashes, domains, IP addresses, exploit paths, malware signatures or forensic samples tied to a specific incident. The available input is regulatory, institutional and comparative, not forensic or operational.

Analysis for Security Teams

For security teams, the Mexican case is more about mapping legal obligations than about hunting threats. The priority is not just technical controls, but also identifying which duties apply in each state, which sectors were affected by labor or copyright reforms, and which AI or cybersecurity clauses are already being pushed down to government suppliers. A company operating across several Mexican jurisdictions cannot assume that one state’s local rule covers the whole country. Yucatán, San Luis Potosí and other local congresses can impose or remove sanctions, change definitions, or leave interpretive gaps. That requires building compliance matrices by territory, by data type and by use case.

At the same time, it is worth separating three layers that the public debate keeps mixing together: generative AI, algorithmic automation and corporate cybersecurity. Mexico’s legislative reaction is focused on deepfakes, voice and image cloning, automated decisions and AI use with minors. The technical response cannot be reduced to a single privacy control. It needs a model inventory, risk assessment by use case, input traceability, controls over vendors and human review procedures where employment, education, service access or administrative processes may be affected.

The cybersecurity front has its own urgency. The 2025-2030 National Plan and the General Policy for the Federal Public Administration create direct requirements for federal agencies and indirect ones for contractors. Companies working with government need to review contract terms, business continuity, incident reporting, ISO/IEC 27001 controls and response programs. In practice, the regulatory perimeter does not stop at the public agency portal. It extends to vendors, integrators, platforms and outsourced services. If there is no general corporate cybersecurity law, the real standard will be the one imposed by public customers and sensitive supply chains.

For organizations with regional exposure, the comparative map also matters. Peru already imposes obligations by risk level and specific deadlines. Chile is moving in the same direction, with risk categories and automated decisions under the umbrella of data protection. Uruguay already requires annual cybersecurity reporting and is expanding national AI capabilities. Paraguay, despite lacking an AI law, faces enough cybersecurity incidents to push reforms in e-commerce, digital identity and response capacity. These are not equivalent countries, and treating them as if they were creates compliance mistakes.

Material Limitations

The file does not include the full text of the Mexican bills, their committee opinions, or consolidated versions of the local reforms. For that reason, it is not possible to reconstruct each state initiative article by article or confirm whether all the budget impact figures have already been turned into operational approvals.

There is also no isolated technical incident that would allow for a campaign narrative with verifiable IOCs. The available material is regulatory and public-policy oriented. As a result, the attack-chain section was limited to a chain of regulatory and operational exposure, not a forensic intrusion.

Several claims in the research are marked by the source itself as uncertain. In particular, some regional comparisons about cybersecurity leadership, or mentions of state attribution in Paraguayan incidents, should be read cautiously. That uncertainty has been preserved in the body of the report.

There is also a timing mismatch. Some sources come from different days in July and August 2026, while others are institutional references without a precise date or with their own internal chronology. The report prioritized confirmed facts and placed less weight on assertions that were not corroborated by more than one source or that were framed as editorial opinion.