In a new ruling, Colombia’s Constitutional Court analyzes the use of AI tools by the country’s Judiciary

In a new ruling, Colombia’s Constitutional Court analyzes the use of AI tools by the country’s Judiciary

In a ruling on August 16, the Constitutional Court of Colombia, the country’s supreme judicial body in matters of constitutional law, examined whether there had been a violation of the fundamental right to due process of law due to a judge’s use of an artificial intelligence (AI) tool to support his decision. In the specific case, it was found that a 2nd instance judge had used generative AI to write a decision granting a request for urgent relief made by a mother, who asked that her child, with special needs, be exempt from paying fees to a health insurance company in order to access medical treatment.

The situation involved an application for injunctive relief, filed in 2023, mainly arguing whether a health insurance company’s demand for co-payments or a fee to authorize medical procedures infringed on the fundamental rights of a child on the autistic spectrum to health and a dignified life. In the second instance, the judge in the case ruled in favor of granting the injunction, but in the middle of his argument, he transcribed excerpts from his interaction with a generative AI tool, which he used to motivate his verdict. In this way, in the Constitutional Court’s view, given the use of AI in issuing the injunction decision, it would be relevant to analyze whether it was adequately motivated or whether it could be the product of hallucinations and prejudices generated by the technology.

With this in mind, the Court sought to cover some of the main points in its judgment, such as: (i) the fundamental right to due process of law; (ii) due process of law in a court system that uses AI; (iii) the AI system: concepts and basic aspects of how it works; (iv) the impact of the use of AI tools on society; (v) the state of AI in Colombia; (vi) AI regulatory frameworks around the world: soft law instruments and national regulatory initiatives; (vii) some concrete experiences related to AI in judicial practice; (viii) the guarantee of the natural judge in a jurisdictional system that uses AI and (ix) the due process of evidence in a jurisdictional system that uses AI.

First, the constitutional court ruled on the possible violation of due process of law due to the fact that the judge had used the AI tool for his reasoning. Here, the Court considered that there was no substitution of AI for the exercise of the judicial function, essentially because the AI system was used after the decision had been reasoned and taken. According to the decision, “in the methodological order adopted by the guardianship decision, the judge first identified the thesis that he was going to support, then the constitutional norms applicable to the case, the jurisprudential reference that should be taken into account due to the factual identity with the matter analyzed, to then resolve the concrete case indicating that the collection of co-payments and moderating fees constituted a barrier to access to the child’s health service and, only then, announced and proceeded to carry out questions in the aforementioned AI system, in order to transcribe the answers given in the consultation”.

Next, the decision analyzes the 2nd instance judge’s compliance with the principles of transparency, accountability and privacy. Here, the court established that the required principles of transparency and accountability were not fully complied with, but on the other hand, the principle of privacy was complied with, since the judge did not enter personal data on the child or his medical history, or on the parties involved in the dispute.

The court then points out the relevance of regulatory instruments to establish a kind of roadmap that, through principles and guidelines, ensures that AI systems are developed and controlled under an ethical approach. As examples, the decision cites the European Charter of Ethics on the Use of Artificial Intelligence in Judicial Systems and their Environment, issued in 2018 by the Council of Europe’s European Commission for the Efficiency of Justice (CEPEJ), as well as the Global Toolkit on AI and the Rule of Law for the Judiciary, published by the United Nations Educational, Scientific and Cultural Organization (UNESCO) in 2023. Both are normative initiatives that aim to provide judicial actors (such as judges, prosecutors, lawyers and law universities) with the knowledge and tools necessary to understand the benefits and risks of AI in the legal environment.

Thus, the Court ruled that, in fact, the right to due process of law had not been violated by the judge in question, and therefore there was no cause for the proceedings to be declared null and void, since the use of the AI tool “did not imply an invasion of the function of the administration of justice by the competent judicial authority”. However, it also noted that, in compliance with constitutional guarantees and fundamental rights, and given the risks of hallucinations, discriminatory biases and others associated with AI, it is necessary to establish guiding criteria, as well as the adoption of guides with guidelines for their implementation by judges, and also for the Judiciary to adopt practices that, in accordance with constitutional principles, allow for the reasonable and proportional use of AI tools, without undermining due process or restricting the autonomy and independence of the Judiciary.

Finally, the Court ordered Colombian judges, in general, to assess the appropriate use of AI tools, considering best practices and applying ethical criteria and respect for current laws, in order to guarantee fundamental rights, especially due process, when they consider it necessary and appropriate to use them, and to ensure judicial independence and autonomy. It also ruled that the use of AI in judicial decisions in Colombia must comply with certain criteria, such as: transparency; accountability; privacy; non-substitution of human rationality; reliability and constant checks; risk prevention; promotion of equality and fairness; ethical regulation; continuous monitoring and adaptation; and suitability. Finally, the ruling ordered the Colombian Supreme Judicial Council to publish, within four months, a guide, manual or guideline on the implementation of generative AI in the Judiciary.

The ruling can be accessed via the link: Sentencia T-323 de 2024

Note: For quick release, this English version is provided by automated translation without human review.