Lazy* police work results in arresting an innocent person.

Aka: The Machine did its job. The lazy* human did not.

https://petapixel.com/2025/08/28/new-york-police-nypd-facial-recognition-tech-blunder-leads-to-62-man-jailed-for-crime-by-56-suspect/ 

Stop Worshipping the Magic Box

Here’s the creeping cancer in modern investigations: treating tools like oracles instead of assistants. At some point, we all do it, right**?

Facial recognition is just a probability engine. Forensic software is just parsing logic. They are calculators, not judges. Even when I use a calculator, I have a habit of calculating two or three times to make sure that I didn’t make an input error. I don’t trust the calculator and I don’t trust myself to be 100% accurate without verification.

So why would we let a software application output stand as evidence without independent confirmation?

The Comfort of Complacency

The dangerous thing about investigative tools is that they’re usually right. Ninety-five or even ninety-nine out of a hundred times, they give you accurate information. That breeds complacency. You stop questioning. You start assuming because the tools are almost always right. I admit, forensic software is amazing, and it is easy to become complacent, especially when it seems the tools get it right almost all the time.

However, in court, it’s not the 95 that saves you that you need to worry about. It’s the five that will damage the case and potentially your career. One unchecked “match” and you’ve framed the wrong guy. One unverified “hit” and you’ve given the defense a gift-wrapped motion to suppress. One misinterpretation of the number of searches made or one lazy reliance on a report, and your credibility takes a direct hit.

The court doesn’t care that the software got it wrong, because it is you who got it wrong.

The Non-Negotiable Rule

In this work, whether it’s digital forensics, incident response, or policing, the rule is the same: verify everything.

The tool is there to give you clues. You are there to verify the clues (prove or disprove, validate, corroborate). Relying on a tool is abdicating your duty.

I bet that if the investigator (DF/IR/PI/LE) were to face prison time should they get it wrong, complacency would disappear overnight….

Case breakdown

A few points on this case, just based on the article alone. Given some free time, I would love to have the full report. Then again, it might be too frustrating to read.

The Not-So-Funny Part

I continually hear “I wish I had this when I first started” with things I’ve written or taught.

That’s a red flag for our field, and any field where your work directly impacts a person, family, or business. It means generations of DF/IR have been thrown into the field without being taught the most basic, non-negotiable rule: verify everything and here is how to do it.

It hurts to realize years later that you were flying blind when you should’ve been trained from day one. But it’s still better late than never. Because the moment you stop feeling that sting of the “I wish I had known this earlier” moment, you’ve stopped learning.

Acting like a bad guy taught me a valuable lesson

I’ve told the long version before, but here’s the short version: After two years working undercover cases (dope, guns, gambling, humans) I was ready to quit. Why? Because the job sucked. It sucked because:

1. I didn’t know what I was doing.
2. My “trainer” was teaching me how to do it wrong.
3. Resources were nonexistent.
4. I was nearly killed, more than once.

At the two-year mark, I took a training that changed everything. It showed me how I was putting myself and others at risk, how to work good cases, and  most importantly, how not to die.

Getting into DF/IR is no different. When you don’t know what you truly need to know, you are dangerous. If you are lucky, you learn early from a mentor or a class, or both.

The points of all this:

AI: It’s a tool. Learn what it is because it will be in everything. Treat it no differently than any other tool in that you must verify what it is telling you.

Training: Don’t neglect taking a class that will make all the difference. A training gap is career and case cancer.

Clues:  Clues are not evidence, probable cause, causation, or correlation until proven and verified. They are tips to check.

Complacency: An easy trap to fall into because your tools are so good.

Shortcuts: Today’s shortcut is tomorrow’s civil lawsuit.

Mentorships: GOOD mentorships matter. I personally know how a bad mentor can risk wrecking your career, so find a good one.

Ethics: Ethics matter in more ways than you can imagine.

————–

*By “lazy”, I mean that although the investigator is capable, competent, trained, experienced, able to do what is required, and they have the opportunity and resources to do it, they choose not to. They either blindly trust their tools, or they simply do not want to get out of their chair or make a phone call to double-check what they assume.

**The answer should be ‘no’ by the way. You didn’t think I’d let you get away with that, did you?