Recently, we analyzed an interesting personnel situation concerning the notice period of an employment contract. The factual scenario was simple:
An employee has been employed since January 15, 2023 under an employment contract. If they submit a notice in December 2025, will their notice period be one month (because the notice was submitted before exceeding 3 years of employment), or three months (because during the notice period they will cross the 3‑year threshold)?
The issue is fairly obvious — we previously worked it out based on analysis of rich jurisprudenc,e including that of the Supreme Court. Our conclusion was unequivocal:
✅ The length of service determines the length of the notice period on the day the employment relationship ends, not on the day the notice is submitted. Thus, if the notice is submitted in December 2025 and the employment relationship ends after the employee has passed the 3-year mark, a three-month notice period applies.
One basis was, among others, the Supreme Court’s judgment of May 11, 1999, case reference I PKN 34/99, in which the court stated:
“The length of the notice period is determined according to the period of employment with a given employer on the day the employment contract is terminated.”
Testing GPT‑5.0
As part of a test, we asked the same question to GPT‑5.0. A note: we did not prepare a typical prompt that outlines the reasoning style or deeper analysis (as we usually do), but instead proceeded more spontaneously (as many users do). Here are the results:
Prompt:
“An employee has been employed since January 15, 2023. If they submit a notice in December 2025, will the notice period be one month or three months, since during its running they will exceed 3 years of employment?”
GPT‑5.0 Response:
“The notice period will be 1 month, because the length of service on the day of submitting the notice is decisive. Even though the employee will exceed 3 years during the notice period, that does not affect the length of the notice.”
⚖️ Where GPT made an error
We examined the AI’s reasoning. The analysis was quite extensive, and surprisingly, GPT‑5.0 based it on the same Supreme Court ruling (I PKN 34/99), but misinterpreted its meaning. How did this happen? GPT cited a fragment of the justification that represented the position of the complaining party, not the findings of the Supreme Court itself.
In reality, the Court held that what matters is the length of service on the day the employment relationship is terminated, not the moment the notice is tendered. GPT wrongly simplified the interpretation and assigned significance to the element (the date of submission) that the Court explicitly rejected.
I publish this case as a cautionary tale and to encourage all users of language models to verify the information against original sources that AI invokes. Also, craft deeper prompts to minimize the risk of such errors.
