# Marcus Delgado, Digital Forensics Examiner at Bernalillo County DA's Office — read of ForensicCheck, June 14 2026

> 9 years doing digital forensics for criminal prosecution, started as a patrol officer, got pulled into cybercrime unit in 2019. Currently the only dedicated examiner in an office of 42 prosecutors. My tools are Cellebrite UFED, Magnet AXIOM, and FTK Imager. I use FotoForensics when I need to sanity-check image metadata fast.

---

## How I got here

My colleague Priya in the cybercrime unit forwarded this link on Teams with just "have you seen this?" We've got two open cases right now involving suspected deepfake video being used in harassment, and finding anything validated enough to survive cross-examination has been a nightmare. So I clicked.

## What I clicked first

"Court-grade synthetic media detection with cryptographic chain-of-custody documentation." That phrase either means something specific or it's the most dangerous kind of marketing copy. In my world, "court-grade" is not a category that exists until a judge says it does. I clicked "Documentation" immediately. Dead link or redirect, not sure which, but it didn't take me anywhere useful.

## Where I paused

The "No Upload History" section. That one line, "Self-hosted or air-gapped deployment available for classified investigations," is actually the right instinct. Whoever wrote this has at least talked to someone in law enforcement, because that's the first objection we'd have: chain of custody over the evidence file itself, not just the report. Uploading evidentiary media to a third-party cloud is DOA with our legal team. So that caught my attention for a second.

## What I distrusted

Three things, and they compound each other.

First: "Identify AI-generated imagery and deepfake video with confidence scoring." What confidence scoring? What methodology? What's the false positive rate on real video? In a courtroom, if your tool flags authentic footage 3% of the time and I can't explain WHY it flagged something, opposing counsel eats me alive. This page gives me nothing to work with.

Second: "Cryptographically sealed reports." Sealed how? SHA-256 hash of the output PDF? A blockchain write? Signed by what key, maintained by whom, auditable how? "Cryptographically sealed" without technical specifics is the same as saying "very secure."

Third, and this one killed it: I scrolled down and found "Adopt this idea" with tiers at $5 and $99. Then I saw the other products listed next to it: "Airline Customer Service AI," "SupplierQ Voice." This is a product concept marketplace. I'm not reading a product page. I'm reading a pitch deck for something that may not exist yet. The "working code starter" language in the $99 tier confirmed it. I'm not buying a code starter for forensic evidentiary work. That's not how any of this works.

## What would convince me

I'd want to see an actual detection methodology with a name I can look up, like ELA analysis, GAN fingerprint detection, or something with a published paper behind it. I'd want error rates on a labeled benchmark dataset, ideally one I've heard of like FaceForensics++ or DFDC. I'd want to know whether any of this has been admitted in a real proceeding, even once, and in what jurisdiction. One actual case reference, sanitized, would do more than every claim on this page combined.

## What I'd ask in an email reply

1. What is the underlying detection methodology, and is there a published accuracy benchmark with a false positive rate on authentic media I can cite in my forensic report?
2. Has any output from this tool been accepted by a court as expert evidence, even in a Daubert hearing where it was argued over and let in?
3. When you say "cryptographically sealed reports," can you describe the specific implementation so I can explain it under cross-examination?

## Verdict: dismissive

The use case is completely real and I'd genuinely pay for something that solved it. But this page is selling a concept, not a tool, and the two claims that matter most in my world ("court-grade" and "forensic rigor") are pure assertion with nothing behind them. I closed the tab.

---
*Memo by skeptic persona, generated 2026-06-14. Studio breaks own self-grading loop.*