Tax 

AI as a tool (not only) in taxation: Helper or risk?

Artificial intelligence can make work easier not only for lawyers and tax advisors, but also for anyone who uses it, for example, to conduct research, review extensive materials or prepare documents. However, the risk of inaccurate or fictitious outputs applies to all users, regardless of their expertise. If an entity submits materials prepared with the help of AI, for example, to a court, authority, bank or subsidy provider without prior verification, the submission of inaccurate or false information and documents can lead to a range of problems. The consequence may not only be, for example, an unsuccessful application or denial of a claim, but also, depending on the circumstances, sanctions or other legal consequences.

A convincing text is not a guarantee of correctness

In addition, AI may misinterpret legal and tax regulations or rely on facts that are not supported by these or other documents. In practice, it can distort, for example, the tax treatment, the content of court decisions, amounts or information on compliance with various conditions. At the same time, the mere reference to the use of AI does noth relieve the user of liability. Any potential sanctions will depend on the specific circumstances.

When AI “hallucinates” – the view of Czech courts

The Constitutional Court (CC) in its order in case no. I. ÚS 3004/25 dated 1 December 2025 addressed the use of AI by legal counsel. In the constitutional complaint, the lawyer relied on a total of 12 court decisions. However, upon verification, the Constitutional Court found that a substantial number of the cited decisions did not exist at all and others had been grossly misinterpreted.

In its order, the CC states that “it is not important with the help of which persons or tools lawyers process a submission to the court”, but at the same time emphasizes the requirement for submissions of an appropriate formal and substantive standard and for legally qualified representation before the Constitutional Court. Subsequently, it imposed a fine of CZK 25,000 on the lawyer for grossly obstructing the proceedings.

Another case was also dealt with by the Supreme Administrative Court (SAC), which in its judgment similarly questioned the content of a cassation complaint. It also cited non-existent, misinterpreted or incorrectly labelled judgments of the SAC, which were supposed to support the cassation complaint. In this context, the SAC used the slang term “AI slop”, which refers to low-quality content generated by AI. The SAC was a bit more lenient towards the lawyer and instead of imposing a fine, strongly urged him to avoid this approach in the future.

The position of the courts is thus quite clear: it is not the use of AI itself that is problematic, but the uncritical adoption of its outputs without adequate verification.

Effective use of AI

However, these risks are not a reason to reject AI altogether. AI can be particularly useful in internal processes, where it can help streamline and speed up routine activities as well as assist with modelling or the preparation of underlying data. We recommend starting with the identification of a suitable process, assessing the benefits and costs of implementing AI, and, above all, establishing rules for working with data and reviewing outputs. We routinely help clients identify suitable areas of use and set up a legal and tax framework for the use of AI. The goal is not to replace specialist teams, but to use technology to make the work more efficient.

What does this mean in practice

For users, it is essential to remember that a convincing AI response is no guarantee of accuracy and does not replace expert assessment of a specific issue. It is necessary to take into account the legislation applicable to the relevant tax period, the actual course of the transactions and other relevant information.

Before submitting a document to a court, tax authority, other public authority, bank or subsidy provider, it is therefore essential to verify not only the legal conclusions, but also the accuracy of the data used and whether they are supported by the original documents. A responsible approach to AI does not mean rejecting it but identifying areas where it can bring real time savings and make work more efficient.

AI Artificial intelligence

Upcoming events

Seminars, webcasts, business breakfasts and other events organized by Deloitte.

    Show morearrow-right