Spain’s Constitutional Court rules out AI for screening amparo appeals

Spain’s Constitutional Court rules out AI for screening amparo appeals

In brief

  • The Plenary of Spain's Tribunal Constitucional unanimously rejected using AI for the initial screening of amparo appeals, as reported on 12 November 2026.
  • AI will be limited to instrumental tasks (case-law search, summaries) in encapsulated environments protecting sensitive data; an in-house system is expected by spring 2027.
  • It is not a judicial ruling, but it sets an institutional line: instrumental automation yes, decision-making automation no.

The Plenary of Spain’s Tribunal Constitucional has unanimously rejected the use of artificial intelligence in the initial screening of amparo appeals, as reported by Moncloa.com on 12 November 2026. It matters because a high court is drawing the line between assistance and delegation before adopting the technology.

What has been decided

According to the report, the rejected request proposed applying AI to the triage of all amparo applications (individual constitutional complaints). The court is studying how to bring in the technology, but only for instrumental tasks: finding case law, drafting summaries and retrieving European Court of Human Rights rulings or scholarly articles. This was explained by Judge Juan Carlos Campo Moreno, who leads the court’s digital transformation.

This is neither a judgment nor a judicial proceeding, but an institutional position of the Plenary. An in-house system, or providers offering sufficient guarantees, is expected by spring 2027, and a committee of legal clerks will decide which tasks are automated.

The rationale: deliberation cannot be delegated

Admission of an amparo requires special constitutional significance under Article 50.1 of Organic Law 2/1979 on the Constitutional Court. The source explains that this assessment belongs to the bench of judges through deliberation, because a new nuance can change the weighing of the whole case. It is tied to Articles 24.1 and 24.2 of the Constitution (effective judicial protection and due process) and to Article 120.3 (reasoned decisions), which presuppose a person weighing the issues rather than an opaque output.

The court rules out black-box systems and names three risks: opaque algorithms, structural bias and blurred accountability.

Sensitive data and verification

The court wants a fully “encapsulated” AI, with guarantees on where documents are uploaded and where data travels. Its clerks also found an amparo application whose references to dates and judgments did not match the decisions cited. The lesson drawn is that the rigour expected of a filing cannot be delegated to a text generator.

Our view at CEA Digital Law

At the Euro-American Committee of Digital Law - CEA Digital Law, we welcome the fact that the court states the limit before deploying the tool. The distinction between instrumental and decision-making automation is easy to communicate and offers a practical benchmark for other jurisdictions, including those in Latin America and the Caribbean and in Africa, where judicial digitalisation is moving quickly and often without an equivalent prior rule.

We suggest three ideas for anyone adapting this approach: first, set out and publish a catalogue of permitted uses; second, require traceability and verifiable human oversight even for support tasks, since a summary also steers a reader; third, train staff in checking citations, given the episode of inaccurate references.

Open questions remain: the border between “summarising” and “sorting” can blur in practice, and no formal implementation timetable has been published so far. It will be worth watching how the catalogue of uses takes shape and whether other courts follow this model.

Sources

CEA Digital Law

النشرة

فعاليات وأخبار وتحليلات في القانون الرقمي تصلك على بريدك، ولا نراسلك إلا حين يكون لدينا جديد.