In recent days, several news reports have emerged about a man and a woman being contacted and summoned by the police based on misidentifications made by facial recognition systems. In both cases, it was later proven that they had not even been at the scene [1, 2].

Now, setting aside the legal background, let’s look at the technological and business lessons. Facial recognition — like any AI-based system — can make mistakes. It is not enough to measure only “how many matches it produces”; the rate of false positives is just as important, as is what happens with those false matches in practice. Where is the decision made? Who is responsible? And what impact does this have on the people involved?

The example is illustrative, but these questions do not apply only to the police. The same dilemmas arise in SMEs as well. Implementing AI on its own is only a partial solution: without organizational readiness, control points, and human decision-making, we can easily end up automating errors too.

AI does not want to make decisions — it is often we who shift that responsibility onto it. In these cases, organizational processes were designed in such a way that the machine’s output appeared as a final decision, tied directly to mandatory action. It is worth asking: how might these cases have unfolded if every match had been reviewed by a human, with that person deciding whether it was truly justified to proceed?

Technology is a powerful tool — but only if we understand its limits and integrate it into organizational processes accordingly.

Sources:

[1] https://index.hu/belfold/2026/02/10/karteritesi-per-arcfelismero-rendszer-teves-azonositas-tolvaj/

[2] Telex: Hungary’s police facial recognition system in action: a woman from Szeged was targeted over the theft of a plush teddy bear in Budapest