Colombia's Supreme Court Sanctions Lawyer for AI-Generated Legal Brief with Fabricated Citations

A lawyer's duty to verify what they submit cannot be outsourced to a machine.
The Supreme Court's reasoning for sanctioning a lawyer who submitted AI-generated false legal citations without verification.

A lawyer was fined 15 minimum monthly salaries for submitting a revision appeal containing AI-generated false legal citations and non-existent court precedents. The AI system created 'hallucinations' with fake jurisprudential references that persisted even after the lawyer's correction attempts, violating duties of accuracy and verification.

  • Lawyer Jorge Hernán Zapata Vargas fined 15 minimum monthly salaries (approximately $5,400 USD equivalent)
  • AI system generated false citations to Articles 250 and 355 of Colombia's General Procedural Code and non-existent court precedents
  • Supreme Court ruling issued November 4, 2025; first judicial sanction in Colombia for AI-generated legal fabrications
  • False citations persisted through lawyer's correction attempt, only acknowledged after second confrontation

Colombia's Supreme Court imposed an unprecedented fine on a lawyer who used AI to draft a legal appeal, which generated fabricated case citations and false jurisprudential references, marking the judiciary's first enforcement action against AI misuse in legal proceedings.

On November 4th, 2025, Colombia's Supreme Court rejected a lawyer's extraordinary revision appeal on formal grounds. But what followed was unprecedented: the court discovered that the brief contained legal citations that did not exist, court precedents that had never been issued, and references to statutes that bore no relation to the case at hand. The source of these fabrications was artificial intelligence.

The lawyer, Jorge Hernán Zapata Vargas, had used an AI tool to draft the appeal in order to speed up his work. The system had generated what technicians call a "hallucination"—confident-sounding but entirely false legal authority. When the court flagged the inconsistencies and gave him a chance to correct them, Zapata submitted a revised brief. It contained new fabricated sources. Only when confronted a second time did he acknowledge that an AI program had authored the problematic passages.

The court's investigation revealed the systematic nature of the problem. One citation pointed to Article 250 of Colombia's General Procedural Code as support for revision grounds, even though that article contains no numbered subsections and addresses evidentiary matters entirely unrelated to the appeal's arguments. Another reference claimed that the grounds corresponded to Article 355 of the same code. A third listed supposed court decisions that could supposedly be verified through official channels—but could not be found anywhere.

This was not a careless mistake or a single slip. The false citations appeared in the original brief, persisted through the correction process, and only stopped appearing once the lawyer admitted the AI's involvement. The court saw in this pattern a violation of the fundamental duty that lawyers owe to the judiciary: the duty to verify what they submit and to speak truthfully to the court.

The Supreme Court did not act in isolation. International courts have already begun sanctioning lawyers for submitting fabricated legal sources generated by AI systems. The court cited these precedents as it decided to impose its own penalty: a fine of fifteen minimum monthly salaries, to be paid to the Judicial Council's fund for modernization and efficiency of the justice system within ten days of the ruling becoming final.

But the decision carried a broader message. The court had already revoked a lower court's ruling in a separate case where a judge appeared to have used AI to draft a sentencing decision—one that mischaracterized evidence and invented details about hearings that never occurred. Now, with Zapata's case, the court was signaling that it would hold lawyers accountable for the tools they chose to use. The judiciary was drawing a line: artificial intelligence could assist legal work, but not at the cost of truthfulness, verification, or the integrity of the record.

What remains unclear is how many other briefs, motions, and appeals may contain similar hallucinations—false citations that sound plausible enough to slip past initial review. The Zapata case is the first formal sanction, but it may not be the last.

The lawyer admitted the errors were a consequence of an AI program he uses to expedite responses and procedures.
— Colombia's Supreme Court decision
In various jurisdictions, sanctions have already been imposed on professionals who submitted briefs containing non-existent legislation or fabricated precedents, and directives have been issued to prevent the incorporation of spurious sources in court proceedings.
— Colombia's Supreme Court, citing international precedent
La Conversación del Hearth Otra perspectiva de la historia
Inventor

Why did the court treat this as such a serious violation? The lawyer could have claimed he simply made a transcription error.

Model

Because he didn't catch it. That's the whole point. A lawyer's job includes verifying what goes into a court filing. When you use a tool that confidently generates false legal citations, you have a duty to check them against actual sources. He didn't.

Inventor

But he did try to correct it when the court flagged the problem.

Model

He tried, but the AI generated new false citations in the correction. That's what made it systematic. It wasn't one mistake—it was a pattern that only stopped when he admitted the AI was involved. The court saw that as showing he hadn't actually verified anything.

Inventor

Did the court ban him from using AI altogether?

Model

No. The ruling doesn't prohibit AI use. It just says: if you use it, you're responsible for what it produces. You have to verify it. You can't pass off hallucinations as legal authority.

Inventor

What happens to the other cases he may have filed with AI help?

Model

That's the question no one's asking yet. This is the first sanction. But if other briefs contain similar fabrications, they could be challenged. The precedent is now set.

Inventor

Is this going to scare lawyers away from AI entirely?

Model

Probably some. But I think the real effect will be different—it will force lawyers to actually use AI carefully, with verification built in. The ones who treat it as a shortcut without checking their work will face consequences.

Quieres la nota completa? Lee el original en El Tiempo ↗
Contáctanos FAQ