Analysis · AI and the law in Colombia

Ruling T-323 of 2024: what every lawyer should know before using AI

HOFUSS · June 2026 · 5 min read

Artificial intelligence has already entered law offices and courtrooms in Colombia. And the Constitutional Court has already weighed in. If you practice law and use — or plan to use — tools like ChatGPT in your work, ruling T-323 of 2024 is required reading. Here is what happened, what the Court decided, and what it means for your practice.

The case that started it

It all began with a tutela (constitutional protection action) filed by the mother of a child diagnosed with autism spectrum disorder (ASD), seeking exemption from copayments for his treatment and coverage of transportation costs. During the second-instance proceedings, the circuit judge used ChatGPT as support: he posed questions related to the case and cross-checked the scope of the answers he received.

The case reached the Constitutional Court on review, and the Court used it to take a substantive position on the use of generative AI in the administration of justice — the first of its scope in Colombia.

What the Court decided

In ruling T-323 of 2024, the Second Review Chamber concluded, in summary:

There was no due-process violation in that specific case, because the judge had already reached his decision before consulting the tool: the AI was used as complementary support, not as a substitute for judicial reasoning.
AI can be a tool — never the judge. The Court was clear that these technologies cannot replace the judgment, evidence assessment or argumentation of the judicial officer.
It flagged concrete risks: hallucinations (false information with a convincing appearance), model biases, and lack of transparency in the reasoning behind decisions.
It ordered regulation: it directed the Superior Council of the Judiciary to publish guidelines on the implementation of generative AI in the judicial branch, which led to the official guide later released by the Judiciary.

The risk the Court flagged is not theoretical

AI hallucinations have already cost careers and cases. In the United States, Argentina and Spain there are lawyers who have been sanctioned or investigated for filing AI-invented case law before the courts: citations with an impeccable appearance, convincing docket numbers… of rulings that never existed. In the words of an Argentine court, delegating case-law research to these systems and using the output "without checking the source" is "extremely risky and even reckless."

For anyone practicing law, the message is direct: the problem is not using AI — it is using it without verifying every citation against its real source.

What it means for your practice: 4 rules

Verify every citation against the source. No AI-generated reference should make it into a filing without you confirming the ruling exists and says what is claimed.
AI assists — you decide. The same standard the Court set for judges: a complementary tool, never a substitute for professional judgment.
Prefer tools that show the source. An AI that answers without citing where the information came from transfers all the risk to you.
Document your process. The transparency the Court demands of judges is good practice for litigators too: know what you consulted and how you verified it.

Our position

At HOFUSS we build on the same logic as T-323: AI is a research assistant, every citation is checked against its real source before being shown, and the decision always belongs to the lawyer. If a reference cannot be verified, we do not show it to you.

Request early access

Sources

This article is informational and does not constitute legal advice. Always verify official sources.