Poland's Top Court Rules Lawyer's Blind Use of AI in Filing Unprofessional

Poland's Supreme Administrative Court (Naczelny Sąd Administracyjny, NSA) has ruled that a professional legal representative acted unprofessionally by filing a complaint written almost entirely by artificial intelligence without checking its content. The filing cited court rulings that either did not exist or concerned entirely different cases.

What the court found

NSA judges had no doubt the filing had been generated by a language model. The court rulings cited in the complaint carried issue dates different from those the representative had provided, and none of them actually addressed the suspension of a tax decision's enforcement, the very issue at the heart of the dispute. In other words, the AI supplied the court with fictitious or irrelevant precedents, and the representative let them into an official filing without verification.

It is another case, increasingly reported in Poland, of so-called AI hallucinations in legal documents, where a language model confidently fabricates facts, quotes or case numbers that do not exist. Here the consequences fell not only on the representative but, above all, on his client, who lost the dispute with the tax authority.

The standard for a professional representative

The NSA stressed that a professional representative is paid for legal services, so their client is entitled to expect filings that meet the highest standards of professionalism. Since AI tools are now accessible to anyone, including non-lawyers, simply copying a model's output without reflection cannot count as fulfilling one's professional duties.

Uncritical use of artificial intelligence to draft court filings by a professional legal representative raises ethical concerns and must be assessed critically - Supreme Administrative Court (Naczelny Sąd Administracyjny), ruling I FZ 104/26

The court noted this is not about banning AI use as such, but about how it is used. Tools built on language models can help draft a filing, speed up case-law research or structure arguments, but they do not relieve a representative of the duty to carry out independent legal analysis and verify every cited provision, ruling or case number.

A growing list of similar cases

Poland is not alone here. In recent months similar cases have already reached courts in the United States and the United Kingdom, where lawyers have faced fines or temporary bans from appearing before specific courts for citing fictitious precedents suggested by chatbots. In Poland, case I FZ 104/26 is one of the first clearly documented rulings in which an administrative court explicitly addressed the quality of an AI-generated filing.

For Poland's legal self-governing bodies, this is a signal that liability for AI-generated content in court filings is no longer a theoretical discussion but a real factor in assessing professional diligence. The Krajowa Izba Radców Prawnych (National Chamber of Legal Advisers) and the Naczelna Rada Adwokacka (Supreme Bar Council) have for some time urged caution when using such tools, reminding members that disciplinary and civil liability toward the client rests with the representative, regardless of whether the mistake was made by a human or an algorithm.

What it means for Polish businesses and law firms

For businesses that use law firms, the case is a practical warning: it is worth asking how a representative verifies AI-assisted content, especially in filings submitted on their behalf before courts or tax authorities. A mistake can cost not just the fee but also the underlying dispute, as happened in the case decided by the NSA.

For law firms and accounting offices themselves, which are increasingly turning to AI for the initial drafting of filings, appeals or tax analyses, the NSA ruling could become a reference point for building internal quality-control procedures. In practice, that means a two-step verification requirement: checking that cited rulings and provisions actually exist, and assessing whether they match the facts of the specific case.

The case entered wider media circulation around the turn of June and July, and in recent days further legal and business outlets have covered it, reflecting growing interest in the question of AI accountability in professions of public trust. Further self-governing legal bodies in Poland can be expected to issue their own guidelines in the coming months on using AI tools to draft court filings.

Sources: Important NSA position. A professional representative should verify whether AI is fabricating facts for them (gazetaprawna.pl), NSA ruling: uncritical use of AI by a representative shows a lack of professionalism (rp.pl)