Monday, July 6, 2026

News

Poland's Top Administrative Court: Careless AI Use in Legal Filings Is Unprofessional

PolicyPatryk RabaJuly 6, 2026

Poland's Supreme Administrative Court ruled that a professional legal representative who uncritically uses AI to draft court filings acts unprofessionally and raises ethical concerns, after finding nonexistent chatbot-generated rulings cited in a filing.

Contents
  1. What the court said
  2. Not just a Polish problem
  3. What this means for Polish law firms

Poland's Supreme Administrative Court (Naczelny Sąd Administracyjny, NSA) has issued a ruling that, for the first time, condemns so unambiguously the uncritical use of artificial intelligence tools by professional legal representatives. The case concerned an appeal against a refusal to suspend enforcement of a tax authority's VAT decision, but it wasn't the substance of the dispute that drew lawyers' attention - it was how the filing itself had been prepared.

NSA judges spotted telltale signs of so-called AI hallucinations in the representative's filing. The chatbot the author had relied on cited rulings that either don't exist in the form presented, carry issue dates different from those given, or address entirely different legal questions than the suspension of a tax decision's enforcement.

What the court said

The NSA stressed that a professional legal representative is paid for the legal services rendered, and their client has a right to expect professionally prepared filings. That standard of professionalism is hard to claim, the court noted, when a representative's work amounts to little more than using a tool equally available to people with no legal training at all.

The unreflective use of artificial intelligence to draft court filings by a professional legal representative raises ethical doubts and must be judged most critically - Supreme Administrative Court, ruling I FZ 104/26

That wording carries weight far beyond a single tax case. As Poland's highest administrative court, the NSA has for the first time addressed, in such blunt terms, a phenomenon that has been building for months within the legal profession: attorneys, legal counsels, and tax advisors are increasingly leaning on language models to draft filings, sometimes without verifying the generated content.

Not just a Polish problem

The Polish case fits into a broader global trend. In the United States, a federal judge in Mississippi threw out an entire lawsuit after both sides submitted filings riddled with chatbot-generated errors. The bots had built their arguments on nonexistent court cases and fabricated rulings, and the judge fined all four lawyers involved between $1,000 and $3,500 each, additionally barring two of them from appearing in court for two years.

In both cases, the mechanism was identical. A language model asked to cite legal precedent produces text that sounds convincing and formally correct but has no basis in actual case-law databases. To a layperson, such text looks like solid legal analysis; to an experienced judge or opposing counsel, the errors surface the moment anyone tries to verify the cited sources.

What this means for Polish law firms

For Poland's legal services market, the NSA ruling signals that courts are starting to actively check whether cited case law actually exists, rather than just evaluating the substance of the arguments. Professional self-governing bodies, the bar association and the chambers of legal counsels, had already flagged the risks of using AI to draft court filings, but until now there was no such clear-cut stance from the country's highest administrative court.

In practice, this means firms using tools like ChatGPT or other legal assistants built on large language models should treat any generated content as a working draft requiring full source verification, not a finished product ready to submit to a court. The risk isn't limited to losing a case; it extends to the representative's professional liability and reputation with the client.

Cases like this will likely multiply as AI tools grow more popular in lawyers' office work, including in Poland, where earlier reports indicated that half of civil servants use such tools informally, without official guidelines. The NSA ruling could become a reference point for other courts weighing similar doubts.

Sources: NSA: A representative may not thoughtlessly use AI when drafting filings (rp.pl), Chatbots took over lawyers' work. A judge annulled the trial and punished the attorneys (spidersweb.pl)

Share: