Colombia’s legal system is grappling with the algorithmic opacity of artificial intelligence tools used in legal practice. The central ruling is the T-067 of 2025, which granted a plaintiff access to the source code of CoronApp, a public pandemic application. This decision was based on the principle that access to public information extends to understanding how state algorithms function, a concept often described as the “black box” problem, where inputs and outputs are known but the internal processing remains obscure.
The court defined algorithmic transparency as the availability of information about algorithmic systems, allowing their operation and performance to be assessed. This transparency has two elements: accessibility, enabling external audits, and explainability, ensuring people understand how the system reaches a result in simple terms. The court’s definition was borrowed from academic studies, highlighting the interdisciplinary nature of the issue.
The Challenge of Defining Transparency Standards
The Joint Directive 007 of 2025, issued by the Attorney General’s Office and the Ombudsman’s Office, adopted this definition for the entire Colombian state. However, it incorrectly presented the T-067 as the original source of the concept, which the court itself borrowed from academic studies. This misattribution raises questions about the legitimacy of generalizing a ruling from a specific case to the entire state apparatus.
The issue lies in the lack of operational details. Neither the court nor the directive specifies the amount of information required, the level of audit needed for accessibility, or the simplicity of language for explainability. This pattern is seen in other rulings, such as the T-323 of 2024, which addressed the use of AI tools like ChatGPT in judicial offices without setting a standard. The T-008 of 2026 further emphasized the need for checks and balances in AI used in public functions, showing the growing recognition of these challenges.
Read Also: Colombia escalates military offensive in Arauca against armed groups
The Complexity of Algorithmic Opacity
Researcher Jenna Burrell identifies three causes of algorithmic opacity: deliberate secrecy, citizens’ technical illiteracy, and the inherent complexity of certain models. Each cause requires a different remedy, but Colombian jurisprudence treats them as a single phenomenon, solvable with a general transparency standard. This one-size-fits-all approach fails to address the subtle nature of opacity, leaving gaps in regulation.
These issues are not entirely new. Proper use of sources, decision motivation, and due process are traditional legal problems amplified by AI. The question is who should fill this operational void, as the current system struggles to keep pace with rapidly evolving technology.
The Search for a Solution
The Congress, with direct democratic legitimacy, could legislate. Judges could decide case by case, but the system has not shown the speed needed to keep up with AI tools. The Executive Branch has acted faster, with the Conpes 4144 of 2025 and the Joint Directive 007, but without the same public debate.
Colombia has not decided which path to follow, and this indecision reflects the practical reality of algorithmic opacity in its legal system. The issue is not the lack of a definition but the need for regulators to understand that each case is different and for officials to know what they are applying. Without practical standards, the legal system risks falling further behind technological advancements.
