These news reports document a disturbing pattern in modern policing: facial-recognition systems are often presented as investigative tools, but in practice a bad algorithmic match can quickly become the basis for handcuffs, jail time, and lasting personal harm. In the publicly known U.S. cases below, people were identified by AI face recognition, that identification was wrong, and police action still moved forward far enough to produce an arrest or detention.

What makes these incidents especially significant is that they were not merely technical glitches corrected quietly in the background. They became real-world wrongful-arrest cases involving lost time, legal costs, humiliation, trauma, and, in some instances, national media attention. Several of the best-documented cases came out of Detroit, where reporting has described multiple arrests after faulty facial-recognition matches, but similar failures have also appeared in other jurisdictions.

Taken together, these articles show that the problem is not only whether an AI system makes mistakes. It is also whether investigators, witnesses, and departments treat a software-generated lead as stronger than it really is. The cases below are useful because they show both the human consequences and the systemic weaknesses behind these arrests.