Poland's supreme administrative court caught a lawyer citing AI-fabricated rulings

What happened

In a Polish VAT dispute, a professional representative filed a complaint asking the court to suspend enforcement of a tax decision, citing several prior court rulings to support the argument. The Supreme Administrative Court (NSA) found that the cited rulings either didn't address the issue at hand or didn't exist as described - fabricated by an AI tool and submitted without anyone checking them.

NSA rejected the complaint and, in its August 2026 ruling (case I FZ 104/26), explicitly criticized the unreflective use of AI tools by a professional representative preparing court filings, calling it a failure of professional standards. The client's underlying case was not helped by the argument - it was actively undermined by it.

The business problem

A professional acting on a client's behalf submitted AI-generated legal citations to a court without verifying them, and the fabrication became part of the public record.

What could have helped

• Professional and legal work carries a verification standard that generic AI use doesn't meet on its own
• A hallucinated citation doesn't just fail to help a case - it actively damages credibility with the court or counterparty
• Human-in-the-loop review isn't optional for anything with legal or regulatory consequence
• Responsibility for AI-assisted work product sits with the professional who signs it, not the tool

Where this points