On 21 September 2026 Justice Minister Juan Bautista Mahiques signed Resolución 469/2026, published in the Boletín Oficial on 23 September. It creates the Mesa Estratégica de Inteligencia Artificial en la Justicia (MEIAJ) inside the ministry's Unidad de Gabinete de Asesores. The table is to help draft a common AI strategy for the justice system. Its tasks are to survey existing tools, identify needs and priorities, assess pilots, and set monitoring indicators. Resolución 469/2026 is a short and modest instrument, and that is mostly to its credit.
The case for caution
The strongest argument for tight rules on judicial AI is that courts are not ordinary AI customers. A flawed summary or a wrongly drafted order can affect someone's liberty or property. The person affected often cannot see how the tool works. Judges may also defer to machine output without noticing. Anyone who wants binding transparency, human-oversight and audit duties before any deployment is making a serious argument, not a hostile one.
Argentina already has some of that structure. Resolución 111/2024 created the Programa Nacional Integral de Inteligencia Artificial en la Justicia. As Marval describes it, the coordinator's annual plan must include impact evaluation, risk control and studies and audits carried out before AI is implemented in any area of the justice system. The new table is told to coordinate with that programme, so it adds a planning layer to an existing safeguard rather than replacing it.
What the resolution does, and what it deliberately does not
Three features stand out.
- It is not a regulation. Article 7 says the strategy will serve planning, coordination and institutional promotion and will not aim to establish a general regulatory regime for AI. The ministry has not drafted a broad AI statute for courts. It is asking officials to find out what already exists.
- It is cheap. Article 9 makes participation ad honorem and states that the measure creates no new structures and no additional budget spending. InfoLEG's text confirms the unpaid basis.
- It has a loose clock. Under Article 10 the MEIAJ must submit a preliminary strategy proposal within 90 calendar days of its first meeting, not of publication. As the SIAP summary notes, no calendar deadline is therefore fixed. We could not find a reported date for that first meeting.
The recitals tie the exercise to EL PACCTO 2.0, the European Union's justice and security cooperation programme for Latin America and the Caribbean, and Article 3 asks the table to facilitate links with it. SIAP reports that a July 2026 working mission with experts from France's national magistracy school helped scope needs. That external technical help matters, because an unfunded table of unpaid officials has little capacity of its own.
The evidence base is already decent
Argentina does not start from zero. The ministry reported in May 2025 on a generative AI pilot in the judiciary coordinated by SAIJ. It involved about 5,000 members of the judiciary, 1,007 test bases and roughly 3,000 prompts. The ministry reported an 81% favourable rating from participants.
Those numbers show real appetite, but they measure satisfaction, not accuracy. The ministry's page gives no quantified time saving and does not name the product. A strategy that treats an 81% favourable rating as proof would repeat the usual mistake of mixing user enthusiasm with verified quality. The table's mandate to assess pilots for usefulness, feasibility, security and impact is the right frame. It will matter only if the indicators it proposes are measurable error rates, audit results and published methods, not adoption counts.
Where proportionate regulation points
From a pro-innovation, pro-rights standpoint, the design choice to avoid a general AI law is sensible. A horizontal statute drafted before Argentina has measured how its courts use these tools would likely be either too vague to bind or too rigid for technology that changes quarterly. Planning through a working group, with pilots and indicators, lets the system learn first and legislate later if it needs to.
The risks are not trivial, though.
- Soft instruments depend on follow-through. A ministerial resolution can be revised or quietly dropped by the next minister. If the strategy never gets published or reviewed, it will have constrained nothing.
- Unfunded mandates stall. The no-spend clause helps the fiscal optics. But pilots, training and monitoring cost money, and the resolution does not say where it will come from beyond possible EU technical support.
- Scope is limited by design. Federal justice is only part of the picture, and provincial courts have their own AI experiments. The table can coordinate and persuade but cannot bind them.
- The resolution does not authorise any tool. According to SIAP, it neither approves a specific system for judicial procedures nor authorises general use of any tool. Judges using AI today still do so without a common standard.
What to watch
Three tests will show whether this is substance or ceremony. First, whether the MEIAJ holds its first meeting promptly and that date is made public, since the 90-day clock hangs on it. Second, whether the draft strategy publishes its indicators and its pilot-assessment method, so outsiders can check the claims. Third, whether it addresses personal-data protection, human oversight and traceability explicitly, as the resolution's commentary suggests it should, rather than leaving them as aspirations.
If the table does those things, Argentina will have a transparent, evidence-led route to using AI where it can cut backlogs, with guardrails proportionate to the stakes. If it does not, the country will have another declaration that sits beside tools judges already use. The lesson for other governments is the sequence: measure first, publish the method, and regulate only the harms the data actually shows.