Mexico's Supreme Court (SCJN) began using artificial intelligence as a support tool for the justices' jurisdictional work on Monday 7 September 2026. The Pleno made the decision. Justice Loretta Ortiz Ahlf confirmed it on 4 September, at a human rights congress at the Universidad de Guadalajara. She said she was not fully in agreement and that regulation of AI use in legal matters should have come first, according to Crónica's report. That report names no tool and describes no safeguards.
The story here is not that a court uses AI. It is that the highest court in the country adopted it while the rules for using it were still undefined.
The strongest case for moving first
The case for the Court's approach deserves a fair hearing. Courts are overloaded, and drafting, summarising, and searching case law are exactly the tasks where language models can save time. A court that waits for a perfect statute may find that clerks and judges are already using consumer chatbots informally, with no oversight at all. Formal adoption can be the more controlled path: it puts the practice in the open, lets the institution choose a tool, and creates a place where accountability can attach. Judicial self-governance is also a legitimate form of rulemaking, and the Pleno does not need Congress's permission to organise its own work.
That argument is credible. It depends, though, on the rules arriving with the tool, and on the public being able to see them.
What is actually known, and what is not
Public reporting on the decision is thin. According to the report, the Pleno decided and one justice dissented in spirit. No tool, vendor, data-handling arrangement, disclosure practice, or review procedure has been published. Ortiz Ahlf's objection is about sequencing: regulation first, deployment second.
The Court's own reasoning matters because it sets the pace for the rest of the judiciary. Mexico's judiciary has just been through a contested elected-judge overhaul. A lower court case shows how quickly the practice can run ahead of the rules. In July 2026 the Tribunal de Disciplina Judicial temporarily suspended Lizbeth González Cortés, head of the Juzgado Tercero de Distrito in Zacatecas. The suspension was notified on 14 July, and Publimetro reported that she is accused of using AI tools to draft rulings, agreements, and other documents. These are allegations, not findings. The same report lists other complaints, including alleged workplace harassment and case backlog, and says the Tribunal has not said whether AI use was a direct cause of the suspension.
That uncertainty is the point. If a judge can be investigated over AI-assisted drafting while the Supreme Court adopts AI officially, the line between permitted and sanctionable use is unclear. A judge cannot tell, in advance, whether a given use of a tool is compliant. A litigant cannot tell whether a ruling was machine-assisted.
What good rules look like
Mexico does not need to invent this from scratch. In December 2025, UNESCO published its Guidelines for the Use of AI Systems in Courts and Tribunals, dated 3 December 2025 and built around fifteen principles. These include transparency, accountability, human oversight, human rights protection, and multistakeholder governance. UNESCO frames AI as assistive, not substitutive, and says it must remain under meaningful human supervision.
Those principles translate into concrete, low-cost rules a court can publish:
- Name the tool and the use. Say which system is deployed, for which tasks (research, summarising, drafting), and which are excluded.
- Keep the judge responsible. Every output is reviewed and adopted by a named human, and the decision belongs to the judge alone.
- Protect case data. Sensitive filings must not flow to external providers without contractual and technical protections.
- Disclose and audit. Record when AI assisted, and review use periodically.
None of this restricts innovation. It is the condition that makes adoption sustainable. Technology introduced without public rules tends to be abandoned after the first scandal, which hurts the case for beneficial legal AI more than careful rules would.
The training gap is the hidden risk
UNESCO's survey of judicial operators in 96 countries found that 44% already use AI tools such as ChatGPT, while only 9% say their organizations have issued guidelines or provided AI-related training. Those figures are global, not Mexican, and the page does not give the fieldwork year, so they should be read as an indication of the pattern rather than a measurement of Mexico. The pattern, though, is the one the Zacatecas case suggests: use spreads faster than governance.
A proportionate response
The right answer is neither a ban nor silence. We would argue for three steps. First, the Pleno should publish the decision, the tool, and the permitted uses, so that the rules are public even if they are provisional. Second, the Tribunal de Disciplina Judicial should state what AI-assisted drafting is allowed, so that discipline turns on clear standards rather than after-the-fact discovery. Third, a short review point, such as six months, should assess what the tool did and did not help with.
Adopting AI as a support tool is defensible. Adopting it without public rules, while a lower-court judge is under investigation for related conduct, hands critics their best argument. Ortiz Ahlf's objection is about order of operations, and that is correct. Rules published alongside the tool would have cost the Court little and protected both the institution and the technology.