Man was arrested and kept in prison because of a facial recognition system failure

Artificial intelligence, which is supposed to help police solve crimes faster and more accurately, is increasingly becoming the cause of ruined lives. Another high-profile case from the United States has once again raised the question: can facial recognition systems be trusted when a person's freedom is at stake?

50 days in jail because of a “similar face”

North Carolina resident Jalil Richardson spent more than 50 days in jail after a facial recognition system mistakenly linked him to a suspect in a car theft in Florida.

The algorithm returned an approximately 85% match with an image from surveillance cameras. That was enough for the police to make an arrest. The key problem was that the investigation effectively relied on the AI result as the main piece of evidence, instead of first checking the alibi.

It later turned out that at the time of the crime, the man was hundreds of kilometers away from the scene and was at work. But by then he had already spent almost two months in custody.

Not an isolated case, but a system

Richardson's story is not an exception. Journalists and human rights advocates note that this is at least one of 14-15 known cases in the United States in which people were arrested or accused because of a mistaken match in facial recognition systems.

In similar cases, algorithms gave a “high degree of match” — 85%, 90% and even “almost 100%” — after which police used these data as grounds for detention.

In many cases, investigators did not check basic facts: the person's location, work schedule, or other objective evidence.

How the error that ruins lives works.

The problem lies not only in the technology, but also in how it is used.

Experts point to several key factors:

poor-quality source images from surveillance cameras

algorithms trained on incomplete or biased data

a high likelihood of errors when distinguishing faces across different ethnic groups

excessive police trust in the “match percentage”

Studies show that such systems can make significantly more mistakes when working with certain population groups, which increases the risk of unjust arrests.

Consequences: ruined lives

Even after charges are dropped, the consequences for those affected remain serious:

— loss of employment and income

— damage to reputation

— psychological trauma

— prolonged involvement in the criminal justice system

In some cases, people spent weeks and months behind bars before proving their innocence.

Human rights advocates are demanding restrictions

The American Civil Liberties Union (ACLU) and other organizations have already filed a number of lawsuits against police and authorities, arguing that the use of AI without strict restrictions violates citizens' basic rights.

Lawyers emphasize that the technology can only be used as a supporting tool, but not as grounds for arrest.

Some US states have already begun introducing restrictions or bans on the use of facial recognition in law enforcement without additional evidence.

The main question is not the technology, but trust

Jalil Richardson's story has become another warning: even the most modern algorithms cannot replace a proper investigation.

And while law enforcement agencies continue to treat AI as “almost proof,” the risk of new mistakes remains high — and the cost of those mistakes is measured in human lives.