Artificial Intelligence in the Colombian Judicial System
The case law framework governing artificial intelligence (AI) in Colombia is largely set forth in Ruling T-323 of 2024 of the Constitutional Court of Colombia, also known as the ChatGPT (OpenAI) ruling.
At that time, artificial intelligence tools such as Claude (Anthropic) and Gemini (Google)—which are better known today—did not yet have the same market presence they do now. However, it is evident that, as of today, the latter two are beginning to gain more traction than the former—that is, ChatGPT. Nevertheless, at the time of the aforementioned Constitutional Court ruling, ChatGPT was the most popular tool, and it was the one used by the Circuit Judge in the appellate proceedings to rule on the decision appealed by the EPS.
The aforementioned ruling established important criteria and guiding principles for the judicial branch regarding the use of generative AI tools by the country’s courts, which is relevant to the topic at hand: artificial intelligence in Colombia’s judicial system.
The aforementioned ruling examined the use of artificial intelligence tools in constitutional review proceedings, in connection with the lawsuit filed by Blanca, on behalf of her minor child, against the EPS.
The Constitutional Court examined the use of artificial intelligence (AI) in decision-making in constitutional protection proceedings, given the use of this tool by the appellate judge, as well as the collection of copayments, the provision of transportation services, and the right to comprehensive treatment for a child diagnosed with autism spectrum disorder, whose mother is classified as living in poverty under the SISBEN system.
The appellate judge in the constitutional protection case, as mentioned, used AI to support the ruling; however, he clarified that the decision he reached was based on his own judgment. Therefore, given the use of this tool in the judicial ruling, the Constitutional Court began to examine whether the judge’s issuance of that decision constituted a violation of the fundamental right to due process, since there was doubt as to whether the decision was issued by the judge or by the generative AI—despite the judge’s clarification regarding the decision.
Therefore, the Court’s Chamber examined the following provisions regarding the use of artificial intelligence in Colombia’s judicial system
(i) the fundamental right to due process
ii) due process in a judicial system that uses AI
iii) AI as a system, concepts of its operation
iv) Impacts of AI use on society
vi) The global regulatory framework for AI
vii) Experiences with the use of AI in judicial practice
viii) the right to a natural judge in a judicial system that uses AI
ix) Due process in evidence in a judicial system that uses AI.
The aforementioned ruling is quite interesting because, in addition to examining the issues mentioned above, it analyzes the rights of children and adolescents as individuals entitled to special constitutional protection, as well as the rights of people with disabilities and the protection of their fundamental right to health.
In that case, it was determined that, while the firm’s use of the AI tool was acceptable, it did not fully comply with the principles of transparency and accountability required for its use.
The judge argued that, in accordance with Law 2213 of 2022—which aims to incorporate information technology (IT)—he decided to use AI to uphold the lower court’s ruling, and to that end, he copied the questions and answers generated by the AI tool ChatGPT into the text of the appellate court’s ruling.
Although, as has been reiterated, the judge in question insisted that he made the decision himself and used AI only as an additional tool—but only after having reached his decision—this clarification is important for the Colombian judicial system’s use of artificial intelligence.
Thus, the Court analyzed the elements of constitutional due process that may be violated by the use of AI:
(i) the right to a judge of one's own jurisdiction
(ii) the guarantee of reasoning in judicial decisions
iii) the guarantee of due process in the presentation of evidence
Thus, the Constitutional Court draws on comparative law to analyze AI, and to that end, it adopts the European Commission’s 2019 definition, which states:
“Software (and possibly hardware) systems designed by humans that, given a complex objective, act in the physical or digital realm by perceiving their environment through data acquisition, interpreting the collected data—whether structured or unstructured—reasoning based on that knowledge or processing the information derived from that data, and deciding on the best course(s) of action to take to achieve a specific objective. AI systems can use symbolic rules or learn a numerical model, and they can also adapt their behavior by analyzing how the environment is affected by their previous actions.”
Similarly, it draws on a more recent concept from the Organization for Economic Cooperation and Development (OECD), which on March 5, 2024, published its definition of an AI system as follows:
"An AI system is a machine-based system that, whether to achieve explicit or implicit goals, infers—based on the input it receives—how to generate outputs, such as predictions, content, recommendations, or decisions, which can influence physical or virtual environments. Different AI systems vary in their levels of autonomy and adaptability after implementation."
The ruling explains that, in an AI system, the data it is fed is essential. Likewise, AI is capable of using machine learning, but one of the most important concepts in AI systems today is generative AI, which uses natural language processing (Large Language Models—LLM—or predictive language systems)—a form of computational linguistics based on human language intertwined with statistical self-learning models—as well as the data it collects or is provided.
Generative AI is a type of intelligent algorithm capable of producing original content in various formats, such as text, audio, video, images, and code.
ChatGPT is one of many generative AI models based on large language models (LLMs), specifically the Generative Pretrained Transformer (GPT). GPT is a machine learning model trained on large amounts of text data to generate coherent responses; however, the ruling notes that this does not mean the information is real, reliable, or verifiable, since LLMs are based on text predictions. ChatGPT, as explained by OpenAI during the Constitutional Court’s review process, is a creative medium or, more specifically, a tool for producing content based on instructions (prompts).
In its terms and conditions, OpenAI emphasizes that its tool should not replace professional judgment, since the generated output is not always correct, much less intended for use in legal, medical, or other contexts that could have a significant impact on people’s lives and decisions. OpenAI stated, “Our service may generate incomplete, incorrect, or offensive output that does not represent the views of OpenAI.”
In addition, ChatGPT states in its terms of use that “users are prohibited from relying on or using the models’ responses: (i) as the sole source of legal or financial advice.”
With regard to ethical principles for AI in Colombia, the following stand out:
(i) transparency and explainability, (ii) human oversight of decisions made by an AI system, (iii) safety, (iv) accountability, (v) non-discrimination, (vi) inclusion, (vii) the primacy of the rights of children and adolescents, and (viii) social benefit.
All of this falls within the scope of these tools, which are mechanisms that support—but can never replace—human work in the administration of justice. Additionally, transparency, data protection, and awareness of biases—derived from understanding the origin of the data used to train the AI—among other concepts, would not be compromised; or, at the very least, the risks would be greatly reduced, since this would be an AI system controlled and operated by and for the judicial branch, which seems like a strong incentive for the executive branch to implement it.
Thus, although the Court did not question the validity of the decision—since it had been made before ChatGPT was used—the fact is that it found that, in this specific case, the principles of transparency and accountability required when using AI tools in judicial decision-making were not fully complied with.
Consequently, in its ruling, the Court sets forth the circumstances under which the principles of transparency, accountability, and privacy are deemed to have been met by legal entities that use AI tools; such entities must comply with these principles if they decide to use information technology tools such as Generative Artificial Intelligence.
At Bustos Abogados (Bustos IP Law), we advise you on the efficient use of artificial intelligence and intellectual property rights.
Contact our expert now:
Henry Bustos
Attorney
LLM in Intellectual Property and Information Technology
