Eight questions, 31 worked cases, anchored to real public records — including synthetic controls built to test the checks themselves. Each case shows what the engine recomputed and which checks fired; where a figure is clickable, it opens the source it came from.
UMFR VERIFY
Double-pledged collateral (rule R2)
Curated demo — real public records
Is the same collateral pledged twice? (the REGISTER thesis)
This is what a verified-instrument registry would catch, demonstrated on public cases — NOT a live cross-lender registry (that needs design-partner tapes; borrower identifiers are stripped from public data under SEC Release 33-9638).
Trust ladderReaches Level 3 of 5 — Cross-Referenced
1Document Present
2Data Extracted
3Cross-Referenced
4Anomaly-Cleared
5UMFR Verified
level not reached — something surfaced for review· this lens does not screen here
levels 1-2 are baseline for any document; the differentiation is 3-5
What UMFR claims and does not claim at each level
L1
Document Present
Claims: The document has been received and stored in the system.
Does not claim: Does not claim the document is authentic or complete.
Liability: None beyond storage.
(baseline for any ingested document (precomputed demo fixture))
L2
Data Extracted
Claims: These values were extracted from the document text by the pipeline, each carrying a source page and quote.
Does not claim: Does not claim the values are correct or the document truthful.
Liability: Extraction quality only.
(baseline — extracted / committed data present)
L3
Cross-Referenced
Claims: The extracted data is consistent across the documents provided and passes the internal recompute / reconciliation checks that ran.
Does not claim: Does not claim the underlying business reality matches the documents.
Liability: Cross-reference accuracy.
L4
Anomaly-Cleared
Claims: No statistical anomalies, fabrication signals, or duplicate-pledge patterns were detected in the data.
Does not claim: Does not claim the absence of sophisticated fabrication that the data patterns may not surface.
Liability: Methodology, not outcomes.
(anomaly surfaced — not cleared (verdict: critical))
L5
UMFR Verified
Claims: The full pipeline ran and every check passed or its warnings were acknowledged; the documents meet UMFR's verification standards under the stated methodology.
Does not claim: Does not claim investment suitability and does not constitute advice.
Liability: Per the disclaimer model.
not assessed(a cross-document reconciliation verdict, not a 0-100 score)
Honest boundary: v1 public-data screening. Anchored to the public record (DOJ indictment, Ch.7 filing) — NOT raw UCC filings. Named parties and the ~$1.4B/$2.2B magnitudes are documented public facts; the per-lender allocation is illustrative. A lien-level red flag corroborated by public facts — NOT independent verification of the per-instrument double-pledge (that is v3, private cross-lender data).
Cross-document collateral reconciliation
Cross-lender double-pledge: the same auto-loan and vehicle collateral pool pledged to three secured parties, summing beyond its value.
Tricolor Holdings (Ch.7, Sept 2025) held ~$1.4B of real auto-loan/vehicle collateral but pledged ~$2.2B to multiple financing sources (a ~$800M shortfall), per the DOJ indictment of its CEO and COO. Documented lenders include JPMorgan Chase, Fifth Third Bancorp, and Barclays.
Public-record corroboration — allegation, not adjudicated
DOJ / trustee corroboration (allegation, not adjudicated): The superseding indictment (US v. Chu et al., SDNY) alleges that Tricolor pledged the same collateral pool — roughly 29,000 auto loans — to more than one financing source, reconstructing ~$2.2B pledged against ~$1.4B of real collateral (a ~$800M shortfall). Trustee counsel described the same over-pledge in the Chapter 7 proceeding. JPMorgan Chase and Fifth Third each disclosed a Tricolor-related charge-off; no per-lender amount is stated here. The ~29,000-loan figure is DOJ-alleged pool composition — a corroboration anchor for the aggregate direction, NOT a per-loan or per-VIN tie-out. This remains v1 public-data screening: an allegation corroborated on the public record, not independent verification of the per-instrument double-pledge.
Verdict: CRITICAL1 exception— a screen, not a score
Collateral pledged to multiple lenders beyond its value
R2 — CRITICAL
R2 (collateral pledged beyond its value)
The Tricolor auto-loan and vehicle collateral pool is pledged to 3 lenders (JPMorgan Chase, Fifth Third, Barclays) totaling $2.2B against a value of $1.4B; over-pledged by $800M.
asset value
$1.40B
total pledged
$2.20B
over-pledge
$800.0M
Evidence:
Per-rule and per-severity discrepancy totals are NOT additive: a single pool can carry overlapping discrepancies across rules (e.g. R3 availability overstatement and R4 cross-aging), so summing them would overstate the amount at risk.
Source: public record
Each citation is a named public source — never a raw UCC filing
Pledge — asserting party
JPMorgan Chase Bank, N.A.
Per-party amount is illustrative of the documented aggregate; role shown, not a sourced per-party figure.
Per the DOJ indictment, Tricolor pledged the same auto-loan/vehicle collateral to multiple financing sources, totaling ~$2.2B against ~$1.4B of real collateral. JPMorgan Chase is a documented lender (DOJ). Per-lender allocation is illustrative of the documented ~$2.2B aggregate.
Provenance: real-anchored-curated@demonstration. Per-lender allocation is illustrative of the documented aggregate; this is a red flag corroborated by the public record, not an independent verification from primary filings.
What we don’t claim
Within a pool, not across lenders — borrower identifiers are stripped from public data under SEC Release 33-9638, so cross-lender matching isn’t possible here, and we don’t pretend otherwise.
The $68.1B map covers public filers and shows the disclosure category implicated in the First Brands collapse — it is not a ‘First Brands detector’; First Brands was private and never filed with the SEC.
Findings are exceptions for review, never accusations.
The VERIFY Score is a triage signal, not a credit rating.
UMFR checks whether reported figures reconcile to the source documents. It does not rate credit, predict default, or recommend investment.
Deterministic in the output. The figures in a UMFR report are recomputed by deterministic logic and traced to source. No language model decides the numbers a reviewer sees.
See it on your filings — as a design partner
UMFR is pre-seed, building with a small number of design partners. If you lend in private credit or invest in the infrastructure, talk to the founder.