News
Poland's Supreme Administrative Court Slams Lawyer Over AI-Fabricated Case Citations
Poland's Supreme Administrative Court dismissed an appeal from a lawyer who cited three nonexistent AI-generated court rulings, calling the conduct a lack of professionalism and opening the door to disciplinary action.
Contents
Poland's Supreme Administrative Court (Naczelny Sąd Administracyjny, NSA) has dismissed an appeal filed by a professional legal representative from Łódź who cited three court rulings in his brief that do not actually exist. Everything points to the rulings having been generated by artificial intelligence, with the lawyer failing to verify the content before submitting the brief to the court.
The case began with a decision by the Provincial Administrative Court in Łódź (Wojewódzki Sąd Administracyjny), which in February 2026 refused to suspend enforcement of a tax authority decision. The taxpayer, represented by a professional legal representative, filed an appeal with the NSA, seeking a suspension of enforcement until the dispute was resolved.
Three Nonexistent Rulings
In the grounds for the appeal, the lawyer cited three administrative court rulings meant to support his argument. The NSA panel checked the case numbers provided and found that none of the rulings actually dealt with the issue at hand, some had been issued on different dates than stated, and their content did not match what the brief's author attributed to them at all.
It's a classic symptom of so-called generative AI hallucination, where a model produces plausible-sounding but fictional facts, in this case case numbers and legal theses of court rulings. Tools such as large language models can confidently generate a quote from a nonexistent ruling if they didn't find matching material in their training data.
A Scathing Assessment
The panel, whose grounds were drafted by Judge Sylwester Marciniak, left no doubt about how it viewed such conduct. The court stressed that using AI tools is not prohibited in itself, but relying on generated content without verification constitutes a breach of professional standards.
Thoughtless use of AI tools by a professional legal representative when drafting court briefs must be judged very critically - Judge Sylwester Marciniak, Supreme Administrative Court
The NSA also pointed to the ethical dimension of the case. The court noted that a client paying for legal services has the right to expect that the brief they receive meets the highest professional standards, rather than being unverified content generated by a publicly available tool the client could have used on their own, without a lawyer's help.
Consequences for the Lawyer
In this particular proceeding, the NSA did not impose a formal penalty, limiting itself to dismissing the appeal as unfounded. The court made clear, however, that citing nonexistent rulings could form the basis for separate disciplinary proceedings conducted by the relevant bar association or chamber of legal counsels to which the lawyer belongs.
This distinction matters: the ruling in the tax case itself is not the forum for determining a lawyer's professional liability, but the court's findings could now be forwarded to the legal self-governing bodies as grounds for opening separate proceedings.
A Growing Problem in Court Practice
The case of the Łódź lawyer is not an isolated one globally. Similar instances of AI hallucinations in court briefs have already been reported in the United States, where courts have fined lawyers and revoked their right to appear in court after they failed to verify chatbot-generated citations. The NSA case, however, is one of the first such clearly documented precedents in Poland.
For Polish law firms and tax advisors, the NSA ruling signals that administrative courts are starting to actively check cited case law for authenticity, not just for the substantive accuracy of the argument. In practice, this means every case number cited in a brief can be checked, and any discrepancy will surface quickly.
In its reasoning, the NSA explicitly noted that it does not dispute the usefulness of AI-based tools as an aid in a lawyer's work. What draws criticism is solely the failure to verify content before filing a brief with the court, that is, a situation in which the language model, rather than the lawyer, effectively determines the shape of the legal argument.

