Forensic Document Review · What You Were Sent
You paid for expertise.
Did you receive it?
Independent forensic review of reports, strategy decks, due diligence and advisory work to determine whether conclusions were reasoned, evidenced and internally coherent — or simply generated output dressed as expertise.
Every delivered report includes cryptographic verification of provenance.
Verify a document
The Problem
The Cost of Mistaking
Output for Analysis
Professional work increasingly looks convincing.
The formatting is polished. The language sounds authoritative. The confidence appears earned.
But in high-stakes environments, the critical question is not:
“Does this look professional?”
It is: “Does this withstand scrutiny?”
The hallucination is rarely the cost.
The decision made from the hallucination is.
We routinely find
→︎
Unsupported conclusions presented with professional confidence
→︎
AI-generated synthesis presented as reasoning
→︎
Citations that do not support the claims they are attached to
→︎
Hidden assumptions embedded inside models
→︎
Fabricated confidence presented as evidence
→︎
Internal contradictions that survived every internal review
Professional formatting is not evidence. Fluent language is not reasoning. Confident output is not analysis.
The Forensic Standard
What a Forensic Review
Actually Examines
01
Evidentiary Integrity
Do sources actually support the conclusions they are attached to?
02
Reasoning Quality
Do claims logically follow from the evidence provided?
03
AI Contamination Review
Was fluent synthesis mistaken for genuine analysis?
04
Internal Consistency
Do assumptions in one section contradict claims made elsewhere?
05
Model & Method Integrity
Can the underlying logic withstand independent scrutiny?
You receive
Forensic Findings Report
Claim-by-Claim Breakdown
Confidence Grading
Evidence Map
Methodology Review
Executive Summary
Optional Review Call
Live Cases
Real Cases.
Real Consequences.
These are not theoretical risks. They are documented examples of generated output being mistaken for verified analysis — and what followed.
Lab 008 · The Rogue Protocol · Gold vs USD Cash
A dataset already included the adjustment the AI was arguing for.
Claim under review
“Gold has dominated USD cash since 1971.” Google AI Mode countered that compound interest frequently matched or outpaced gold.
The AI argued for a T-bill reinvestment adjustment that had already been applied to the full dataset. Confident synthesis directly contradicted by the supplied evidence. After forensic review: gold wins across every tested period.
Failure mode
Confident synthesis contradicted by the evidence it was given.
Read full investigation →︎
Case · West Midlands Police · AI in Operational Policing
Fabricated intelligence entered an operational policing report.
What was presented
AI-generated content submitted as verified intelligence within an active operational briefing document.
Parliamentary scrutiny, institutional fallout and public correction followed. The failure was not in the AI’s capability — it was in the institutional assumption that generated output had been verified.
Failure mode
Generated output mistaken for verified intelligence.
Case · Sullivan & Cromwell · Emergency Court Filings
Hallucinated legal authorities entered emergency court filings.
What was presented
AI-generated legal citations submitted as valid authority in formal court documents.
Public apology and judicial embarrassment. Elite institutional safeguards — the peer review, the seniority, the reputational stakes — still failed to catch unverified output before it reached the court.
Failure mode
Elite safeguards still failed to catch unverified output.
Case · Pinsent Masons · Legal Work Product
AI-assisted work product passed internal review without verification.
What was presented
AI-generated legal analysis presented to clients as completed work product without evidentiary verification.
The failure was structural. The process did not distinguish between work that had been generated and work that had been verified. The outputs were indistinguishable in format — but not in integrity.
Failure mode
Format indistinguishable from verified analysis. Integrity was not.
How It Works
Six Steps.
One Forensic Brief.
Step 1
You Submit
the Report
the Report
Step 2
Forensic
Examination
Examination
Step 3
Evidence &
Reasoning Audit
Reasoning Audit
Step 4
AI Contamination
Review
Review
Step 5
Forensic
Findings Brief
Findings Brief
Step 6
Cryptographic
Verification
Verification
Before You Commission
See Exactly
What You Receive
Before commissioning a review, examine a completed forensic report. The structure, the claim-by-claim breakdown, the confidence grading — all visible before you commit.
People buy deliverables. The sample report removes every question about what the engagement produces.
Claim 001 · Evidentiary Review
Claim 002 · Reasoning Audit
Claim 003 · AI Contamination Review
Pricing
Transparent. Fixed.
Quoted before work begins.
No open tabs. No hourly billing. A fixed fee is quoted after intake — before the forensic review begins. £500 intake fee credited in full if you proceed.
Forensic Review
Begin — £500 Intake →︎
Standard Engagement
From £2,500
£500 intake · credited to full review
Ideal for
Investment theses · Advisory reports · Due diligence packages · Internal strategy papers
Includes
Forensic findings report
Claim-by-claim breakdown
Confidence grading
Evidence review & contamination assessment
Executive summary
SHA-256 cryptographic verification
Deep Review
Book a Discovery Call →︎
High-Stakes Submissions
Custom Scope
Scoped individually after discovery call
Designed for
Larger documents · Litigation-adjacent material · Complex multi-part submissions · Boards and investment committees requiring documented independent verification
Includes
Everything in Forensic Review
Extended methodology review
Structured evidence map
Optional findings call with written record
Priority turnaround available
Common Questions
Before You Commission
Why not use GPT detection tools?
Detection tools identify whether text may have been generated. They cannot determine reasoning quality or evidentiary integrity — which is where the real risk sits. A document can be entirely human-written and still contain fabricated citations, unsupported conclusions and hidden assumptions. The forensic question is not who wrote it. It is whether it holds up.
Isn’t AI just another research tool?
Yes — when verified. The issue is not AI’s existence. The issue is unverified output presented as completed analysis. The format is now indistinguishable from expert work. The standard of evidence is not.
Who commissions reviews?
Family offices, operators, advisors, legal teams, trustees and high-stakes decision-makers who have received work product they are expected to act on — and who need independent confirmation that the reasoning and evidence hold up before they do.
What is AI contamination?
AI contamination occurs when generated synthesis is presented as — or mistaken for — independently verified analysis. The output looks like reasoning. It reads like expertise. It does not carry the evidentiary standard of either. Contamination review identifies where this has occurred within a document and what claims are affected.
The Rogue Protocol · Forensic Document Review
Before acting on high-stakes work,
verify what you received.
Professional formatting is not evidence. Fluent language is not reasoning. Confident output is not analysis.
