Audience protection or government overreach?
Mexico’s telecoms regulator (CRT) has opened a public consultation and a strident debate about freedom of expression with new draft guidelines on audience rights.
The guidelines, stemming from the telecommunications and broadcasting law passed in 2025, would require television and radio broadcasters, cable operators and programmers to protect audiences’ right to truthful and contextualised information. This would entail drawing a clear line between news and opinion; disclosing advertising and sponsorship; adopting a code of ethics addressing audience rights; and appointing an independent audience ombudsperson, who must resolve complaints within 20 working days. The draft also tells broadcasters they may not present historical material as though it were current, a clause Leopoldo Maldonado of Article 19 reads as cover for the previous administration against reporting built on the publication of historical archives.
Print media and digital platforms remain outside the regulation’s scope (for now).
The IFT, the constitutionally autonomous regulator the CRT replaced, had published its own audience rights guidelines in 2016, and Congress cut them back by reform the following year. The Supreme Court struck that reform down and handed the regulator’s rulemaking powers back, after which the IFT approved a fresh set of Lineamientos Generales in December 2024. Those are still in force. What the CRT has put out for consultation, then, writes over a framework that already exists and that has already survived a trip through the courts, which gives the industry a reasonable question to ask about what the new text adds.
The legal counsel to the federal executive, Luisa María Alcalde, presented the guidelines as an initiative to reduce discretionary legal interpretation and give audiences concrete enforcement mechanisms, characterising the arrangement as self-regulation (broadcasters appoint their own ombudsperson). In some cases, however, the CRT can intervene directly and impose sanctions.
For example: if a sponsored segment, a paid interview or a branded integration that isn’t clearly flagged as commercial content triggers an audience complaint, and the ombudsperson confirms a violation that the broadcaster fails to remedy, or that the complainant disputes, the CRT can fine the broadcaster up to 1 per cent of its accumulable revenue.
Article 53 of the draft is where the self-regulation framing looks questionable. It says the Commission will supervise compliance by ombudspersons, open television and radio concessionaires, pay television and audio concessionaires and programmers, and it allows monitoring, formal requests for information and whatever other action the Commission considers necessary. No audience complaint is required to set any of that in motion.
Irene Levy, president of Observatel, calls this the provision that worries her most, because it lets the Commission supervise output and open sanction proceedings without anyone having asked it to step in, and she finds no basis for it in either the law or the Constitution. Salvador Camarena, writing in El Financiero, landed on a phrase that has stuck: whether or not this is censorship, it is media policing.
The National Chamber of the Radio and Television Industry (CIRT), which represents more than 1,200 affiliated stations, has said it supports the underlying aim of audience protection but wants tighter definitions of ‘false’ and ‘decontextualised’ information, arguing that as currently drafted these terms leave too much room for subjective interpretation. The association also objects to letting the CRT act as an appeals body over ombudsperson decisions and has flagged the revenue-based fines as disproportionate.
Levy goes further. She argues that the CRT is expanding its authority beyond what is provided for in the telecommunications law, and that such powers should not be set by administrative guidelines. She has also questioned the process by which the draft was unveiled: it was presented by the executive’s legal counsel rather than by the CRT itself, which drafted the document, a sequencing she says undermines the technical autonomy the new regulator is supposed to have.
That autonomy is itself doubtful, because the CRT does not have the standing its predecessor did. The IFT was a constitutionally autonomous body. The CRT sits inside the Agencia de Transformación Digital y Telecomunicaciones, which reports to the presidency. Maldonado describes the result as the government acting as judge and party, deciding for itself what counts as true and then sanctioning broadcasters that fall short of its answer. For a company weighing where to place branded content, the relevant point is who will eventually rule on whether that placement was disguised advertising, and how far that person sits from the executive.
Opposition legislators and press organisations argue that terms such as ‘sufficient context’ give the regulator too broad a degree of discretion over editorial decisions. Other critics say giving the state the job of deciding where fact ends and opinion begins makes self-censorship the easiest response inside a newsroom, particularly in a country where reporting can already be subject to undue government pressures.
The CRT’s public consultation portal is open to comments from concessionaires, companies, academic institutions, civil organisations and individual citizens through 21 August.
Mexico doesn’t have an AI Law. But regulation is happening anyway
There is no AI-specific statute in Mexico, and a recent study, authored by Narciso Campos, Eduardo Flores and Mónica Andonegui of top U.S. law firm White & Case, lays out why that gap hasn’t stopped enforcement. AI gets regulated the way most things get regulated: through whatever existing framework happens to touch it first.
Data protection laws catch it when personal information is involved; consumer law catches it when a chatbot gives bad advice; labour law catches it when a company automates a layoff. Industrial property law now explicitly extends infringement liability to conduct ‘carried out through AI systems’, meaning the legal exposure isn’t waiting for a new AI law — it just needed someone to point an existing one in the right direction. The Supreme Court did its part in July 2025, ruling that AI-generated content cannot hold copyright because authorship requires human individuality and feeling; a decision that reads less like a technical ruling and more like a philosophy exam (with real commercial consequences).
The legislative pipeline tells its own story about how much appetite actually exists to formalise this. Citing Pedro Salazar, a constitutional law scholar at UNAM’s Instituto de Investigaciones Jurídicas, the authors note that roughly 200 AI-related bills have circulated, about 70 remain pending, and as of April 2026 exactly one had meaningfully advanced: a narrow amendment protecting voice actors and dubbing professionals from unauthorised AI voice cloning. Three competing frameworks for a general AI law languish in Congress: one built around risk tiers and mandatory labelling, one built around a dedicated regulatory agency, and one requiring a constitutional amendment before Congress can even legislate on the topic at all. None has advanced to a vote.
In the absence of legislation, soft law and case law are filling part of that gap. In January 2026, SECIHTI and the Agency for Digital Transformation and Telecommunications issued the non-binding ‘Chapultepec Principles’, an ethical declaration meant to guide AI policy across the public sector without creating enforceable obligations. That same month, the judiciary went further: a federal court issued binding precedent requiring that AI tools used in judicial proceedings meet minimum standards on transparency, explainability and human supervision, with judges obliged to disclose any AI use and treat it strictly as an auxiliary tool rather than a substitute for deliberation.
What’s actually enforceable is data protection, and the White & Case team makes the case that it is carrying more weight than a privacy law was built for. The Federal Law on the Protection of Personal Data Held by Private Parties doesn’t define AI, doesn’t require impact assessments, and doesn’t mandate algorithmic auditing, but it does require consent, purpose limitation, and reinforced protection whenever processing involves sensitive data. This is a category that keeps expanding as courts and regulators decide that location tracking, transaction patterns, and biometric identifiers reveal more about a person than the categories originally anticipated.
Mexico’s public-sector AI plan
On the subject of AI, a Mexican government artificial intelligence plan dated April 2026 but posted by the ATDT more recently, sets out how the State intends to use AI across the public sector.
The title promises something wider than what the document covers. Its subject is the government itself: public services, administrative capacity, regulation, official data and technological infrastructure. Helping Mexican companies build commercial models or compete with the global technology groups is not the aim.
The plan argues that Mexico should develop enough internal expertise to avoid becoming merely a customer of systems built abroad. It rests on the aforementioned Chapultepec Principles, ten criteria published on 29 January 2026 by the science ministry and the ATDT, which require explainability, human responsibility and accountability when algorithms shape public decisions. One of them says that a decision which cannot be explained should not be automated. Those principles matter particularly in areas such as tax enforcement, customs inspections, access to government services and emergency management, where errors or biased data can have direct consequences for citizens.
Some of the programme is already operational. A 100-person “AI Factory” is developing tools for government agencies, and its first year produced six of them: a customs risk model for the national customs agency, a conversational guide to federal procedures that has reached more than 280,000 users of gob.mx since January, a bilingual tourism assistant for the tourism ministry, patent-classification software for IMPI, a graph model that maps the networks of companies behind fraudulent invoices for the tax authority, and a system that consolidates reports of missing, hospitalised and dead people during emergencies. The public training centre feeding that talent pipeline runs with 14 technology companies and put 10,000 students through its first intake, with 25,000 certifications a year projected.
The most expensive element is infrastructure. Coatlicue, a proposed publicly owned supercomputer budgeted at 6 billion pesos and developed with the Barcelona Supercomputing Center on the IPN campus at Zacatenco, is meant to reach 314 petaflops across 14,480 GPUs over a 24-month build. It is intended to support scientific research, climate modelling, agriculture, energy planning and the analysis of fiscal and customs data. A broader network of government data centres, anchored on Aguascalientes and a second site at Tulancingo, would provide storage, cloud services and greater control over official information.
The plan offers a reasonably coherent vision of AI for the state, and should be commended. But the devil will be in execution. It contains few measurable deadlines, and there is likely not the budget nor the expertise to implement many of the proposals. If the government struggles to build trains, whose underlying technology has not changed that much in 200 years, then it would take an optimist to see Mexican public sector leadership in AI. (To be fair, the UK, a leader in AI, has also struggled with trains, cancelling most of the HS2 project after spending and in good part wasting so far GBP45bn.).
Setting a new boundary for mining
On 20 July, SEMARNAT published an agreement that settled a longstanding argument. For years, legacy concession holders inside the country’s 232 protected natural areas, covering roughly 98 million hectares, argued that a pre-2023 concession title preserved a viable path to environmental clearance. The Supreme Court spent 2025 and 2026 quietly dismantling that argument, confirming that a concession grants the right to claim a deposit, not the right to dig it up. This agreement is the administrative period at the end of that judicial sentence: it leaves CONANP, the conservation agency, with binding authority over any environmental permit request affecting Protected Natural Areas, rather than the advisory role SEMARNAT could previously weigh against other evidence.
Back in February, the Ministry of Economy disclosed that 1,126 mining concessions, covering 889,512 hectares, had already been returned to the federal government, and that 713 of them (spanning approximately 249,108 hectares) sit within Protected Natural Areas. July’s agreement formalises the environmental-permitting consequence of a clean-up that had been underway for months. On the investment side, Camimex puts exploration spending at $543 million in 2023, dropping to $500 million in 2024 and a projected $400 million in 2025, a decline of roughly 26 per cent that the administration can frame as market discipline rather than deterrence.
The Vulcan Materials case is the headline-making backdrop. Vulcan’s Mexican subsidiary, Calica, operated a limestone quarry near Playa del Carmen for decades, extracting material for construction markets on the US Gulf Coast. Mexican authorities shut down parts of the operation between 2018 and 2022, citing environmental damage to the surrounding karst and coastal ecosystem, and the government eventually designated the area a protected natural zone, permanently foreclosing any return to extraction. Vulcan turned to ICSID, the World Bank’s investment arbitration body, arguing the shutdown amounted to an indirect expropriation under NAFTA. On 27 July, the tribunal dismissed nearly all of Vulcan’s claims, upholding a violation only for the single January 2018 closure, which Mexico’s Ministry of Economy says represents under 1 per cent of the damages originally sought (cited variously as roughly $1.5 billion to $1.7 billion, depending on the source). A government source told Reuters the resulting payment would total around $15 million; Vulcan itself has called the award “insignificant” without disclosing a figure, and the tribunal’s full decision remains confidential for now.
What Vulcan actually tested was whether international arbitration would treat Mexico’s environmental permitting decisions as a violation of property rights. The tribunal’s answer, in effect, was no: shutting down a mine for environmental reasons isn’t expropriation just because the company holds a valid concession, and arbitrators aren’t going to second-guess that judgment except in narrow, specific instances. That’s the real risk for mining companies now operating in Mexico: a concession no longer guarantees a path to extraction.
The Court comes back from vacation
The SCJN returned from recess on 3 August with more than 70 matters queued for its first week, and buried in that docket sits the IMMEX double-VAT dispute. The SAT is chasing roughly $16 billion MXN across 30 lawsuits filed by 24 companies, with Samsung alone accounting for $6.7 billion of that.
This case was inherited from the previous court, and every month of delay is a month that 5,241 IMMEX manufacturers spend not knowing whether the tax framework underpinning nearshoring investment actually works the way they were told it does. Industry groups have been explicit that a ruling favouring the double charge would land as a negative signal for foreign investment, precisely when Mexico is trying to sell itself as the nearshoring alternative to Asia.
Mandatory pretrial detention, subject to a binding Inter-American Court order requiring its elimination, remains unresolved with no date attached. Cannabis self-consumption permits are back for another round, this time testing whether COFEPRIS can simply deny an application without the Court forcing the issue. Other cases include the legality of the definitive export of a Frida Kahlo work, and whether Interjet’s former owners personally owe $689 million pesos in unpaid taxes.
Chatter Box
Morena’s new candidate filter. Morena is beginning to narrow the field for the eight governorships it will contest in 2027, and the process increasingly looks less like a traditional political contest than an exercise in risk management. The party registered 277 aspirants nationwide and is now reviewing their files before deciding who will advance to the polling stage in September. Recent closed-door meetings with contenders from Baja California, Baja California Sur, Chihuahua, Michoacán, Nayarit, Sinaloa, Sonora and Tlaxcala suggest that electoral competitiveness remains important but is no longer enough on its own. As the respected security analyst Eduardo Guerrero notes in a recent column, candidates must also survive integrity checks, restrictions on premature campaigning and, according to reports surrounding the process, scrutiny involving tax, financial-intelligence and security authorities.
Sinaloa provides the clearest indication of how that filter may work in practice. Enrique Inzunza and Juan de Dios Gámez, both previously plausible succession figures, were among ten current and former officials charged by US federal prosecutors in April in a case alleging protection of Los Chapitos; the accusations remain unproven and both retain the presumption of innocence. Neither appears among the twelve publicly identified subsequent registrations for Morena’s state process. That does not establish that Washington is choosing Morena’s candidates, and reports of similar US-related vetting in Baja California and Baja California Sur have not been publicly corroborated. It does, however, demonstrate the political problem facing the party: nominating someone exposed to a US indictment, sanctions, visa cancellation or an extradition request could turn a state campaign into a bilateral crisis overnight.
The emerging model therefore appears to be a graduated filter rather than a blanket exclusion of anyone facing controversy. Figures carrying electoral complaints, political accusations or investigations that ended without charges remain in contention, while risks capable of threatening the party nationally appear to carry considerably greater weight. Polling will still matter, as will gender parity, governors, internal factions and negotiations with coalition partners PT and PVEM, but the pool reaching the survey stage may already have been substantially shaped beforehand.
Transparency after the INAI. President Claudia Sheinbaum has promised another adjustment to Mexico’s post-INAI transparency system, with Anti-Corruption and Good Government Secretary Raquel Buenrostro due to present a presidential decree and proposed legal changes on Tuesday. Sheinbaum said government information should be more readily available through transparency platforms, and suggested that information should only be withheld where justified by national security or ongoing judicial proceedings, with personal data remaining protected throughout. The announcement is potentially significant, coming little more than a year after the former autonomous transparency authority was dismantled and its responsibilities redistributed across Transparencia para el Pueblo, the Anti-Corruption Ministry, and the internal guarantor bodies of other branches and institutions.
Exactly how far the government intends to go will depend on the texts Buenrostro presents. The transparency law adopted in March 2025 currently permits information to be temporarily classified for reasons of ‘public interest or national security’, meaning Sheinbaum’s narrower formulation could amount to a substantive restriction on the grounds available to government agencies for withholding information. That the government has announced both a presidential decree and a legislative proposal raises the possibility of a two-track approach: immediate administrative changes within the federal Executive, alongside broader amendments requiring congressional approval. Equally, it could prove less ambitious, focusing instead on better justification of reservations, public versions of documents, and the amount of government data actually uploaded and searchable through the National Transparency Platform.
For companies dealing with government, the details could matter considerably. A stronger presumption of disclosure could expose more contracts, regulatory files, technical studies, inspections, and information submitted by private parties, making the distinction between genuinely protected commercial information and material merely labelled confidential increasingly important. The political question is equally significant: having eliminated an autonomous body partly on the argument that transparency could be delivered more efficiently from within government, the Sheinbaum administration now needs to demonstrate that the replacement system can produce greater access without giving the Executive excessive discretion over what remains hidden. Buenrostro’s presentation should provide the first real indication of whether this is an operational repair to the post-INAI model, or a more substantial rewriting of Mexico’s transparency rules.
Seven committees walk into a hospital. On 20 July, the Ministry of Health published the organising rules covering seven Specialised Inter-institutional Committees, splitting responsibility across operational, technical, financial, digital, legal and evaluation functions among IMSS, ISSSTE, IMSS-Bienestar, CCINSHAE, Pemex’s hospital network, federal hospitals, and any state health services that opt in. Sitting above the seven committees is a Grupo Estratégico y Coordinador, chaired by the Ministry of Health, which reviews and approves whatever the committees propose.
The 20 July publication sets up who talks to whom and who signs off on what, but it doesn’t yet contain tariff values, a service catalogue, compensation procedures between institutions, or access guidelines for patients trying to use a facility outside their own institution. Coordination on paper and coordination in practice tend to diverge exactly where the money and the systems live: IMSS and ISSSTE have separate financing structures, separate digital health records, and separate union relationships with their own negotiated protections. The rules bind these institutions into a shared structure, but binding language and functional integration are not the same accomplishment.
SAT tweaks digital tax rules. On 9 July, the SAT updated its 2026 tax rulebook, the Resolución Miscelánea Fiscal. Most of the update is routine housekeeping (SAT vacation dates, paperwork fixes for scrap dealers and antique sellers) but four changes affect foreign companies that sell digital services in Mexico such as streaming, apps, marketplaces, betting platforms, and each comes with a real consequence.
Airlines get a break; platforms get less paperwork. Platforms that sell flights on behalf of airlines no longer have to withhold ISR or IVA from those airlines, provided they meet the resolution’s conditions.
Foreign digital providers now follow revised steps to obtain their e.firma, the electronic signature Mexico requires for tax purposes, and some digital providers must now report how many transactions they process each month, not just how much revenue they made. If your transaction count and your reported revenue don’t tell a consistent story, that’s now a visible mismatch instead of an invisible one — this is the change most likely to trigger a question from the tax authority. Online betting and raffle platforms also get explicit rules for calculating IEPS, the special production and services tax, based on money actually collected. With new offshore betting platforms like Polymarket and others gaining ground in Mexico, the tax burden on local gambling firms is only to encourage that migration.
Cybersecurity math. PwC’s 2026 Digital Trust Insights survey found that almost half of Mexican organisations reported their worst breach in the past three years cost somewhere between $100,000 and $10 million, yet only 40 per cent actually quantify that financial exposure before deciding how much to spend defending against it. That gap, between knowing a breach is expensive and calculating exactly how expensive, is where the ‘invest now’ logic comes from: budgeting against an unquantified risk that everyone already agrees is large.
AI-powered malware tops Mexican executives’ concerns at 69 per cent, ahead of AI-enabled supply chain attacks (62 per cent) and deepfake-related risk (51 per cent), meaning the threat model has shifted faster than most compliance frameworks have caught up. Cloud security absorbs the largest share of new budget (46 per cent), and 57 per cent of executives say they’re now spending more on prevention than reaction.
Contacto:
Laura Camacho
Directora Ejecutiva de Asuntos Públicos de Miranda
laura.camacho@miranda-partners.com
Descargar PDF: MI Public Affairs Chatter 260803 – ENG MI Public Affairs Chatter 260803 – ESP