When AI Can't Tell Us Apart, Who Pays the Price?

I ran a simple experiment. I uploaded the same photo of singer Durand Bernarr into three different AI apps and asked each one to identify him (because I'm old and I didn't know who dude was on the awards show I was watching).

The results?

One confidently said Usher. Another insisted it was Lucky Daye. The third... well, it found another random Black man that wasn't even close.

Three guesses. Three wrong answers. Three completely different people.

If AI can't correctly identify a Grammy-nominated artist whose photos are all over the internet, why should anyone trust it to identify the average citizen standing in front of a camera?

Courtesy of 2026 BET Awards

Now let's talk about the elephant mounted on every telephone pole.

Flock cameras.

They're popping up everywhere. Neighborhoods. Shopping centers. Apartment complexes. Highways. The sales pitch sounds great: safer communities through smarter technology.

Who doesn't want safer neighborhoods?

But there's one question that doesn't get asked nearly enough.

What happens when the technology gets it wrong?

History has already shown that facial recognition technology has often performed worse on people with darker skin tones than on white faces. Researchers have documented those disparities for years. While many systems have improved, they're not perfect, and not every company uses the same technology or the same standards.

That matters.

Because if an AI system mistakes one Black man for another, the consequences aren't just an embarrassing error message.

• Someone gets questioned.
• Someone gets detained.
• Someone has to prove they weren't where the computer claims they were.

And once suspicion enters the room, it doesn't always leave just because the facts finally show up.

My little experiment wasn't a scientific study. It was simply a reminder that confidence isn't the same thing as accuracy. The AI didn't hesitate. It answered as if it knew exactly who it was looking at.

It didn't.

That's what concerns me most.

Computers don't have to be malicious to cause harm. They just have to be wrong often enough in situations where being wrong carries serious consequences.

We're racing to install cameras everywhere while assuming the software behind them deserves our trust.

Maybe it does. Maybe it doesn't.

But shouldn't we demand proof before allowing algorithms to help identify suspects?

Technology should assist investigations, not replace common sense.

If AI can confidently mistake Durand Bernarr for Usher, who's to say it won't confidently mistake your neighbor for someone wanted three counties away?

The danger isn't that AI makes mistakes. People do that every day.

The danger is that people often believe a computer simply because it's a computer.

And that's exactly when common sense needs to be louder than artificial intelligence.