The AI TMS Blueprint · Sheet 03

Carrier vetting and freight fraud screening

What each federal authority actually proves and what it quietly does not, how the schemes that are taking money right now actually run, and the control categories that stop each one. Written from the statutes, the regulations and the agency's own words.

Book a 30-minute call with our expert The rest of the blueprint

What changed in 2026

Your vetting file is now evidence

In May 2026 a unanimous Supreme Court held that the FAAAA's safety exception saves a state-law negligent-hiring claim against a freight broker from preemption — requiring a broker to use ordinary care in selecting a carrier “concerns” motor vehicles.

The pleaded basis for the broker's constructive knowledge was the carrier's Conditional safety rating. What you checked, and what you recorded, is now the record a court reads.

Montgomery v. Caribe Transport II, LLC, No. 24-1238

May 2026

A unanimous Supreme Court let negligent-selection claims against brokers proceed

$750,000

Federal public-liability floor for general freight — not the $1M usually quoted

11,671

Compliance reviews in FY2019 against a carrier population above 567,000

$4,500

Fixed-price pilot: your onboarding flow, three weeks, instrumented

The flow

Eight stages, and vetting is a subscription, not a gate.

An insurance filing you read as current can be thirty-five days from void, because that is the notice period for cancellation. Anything that treats vetting as a one-time check is already wrong by design.

01 · judgementCapture the claimSomeone asserts they are a carrier. Nothing has been verified yet — at this stage it is a claim arriving down a channel they chose.
02 · federal lookupResolve to a registrantMatch the assertion to a federal registration record by USDOT number. The number, not the MC docket, is the identifier the agency builds on.
03 · federal lookupAuthority status and typeActive, inactive, revoked or suspended — and whether the authority covers this movement. Property is not passenger, and carrier authority is not broker authority.
04 · judgementBind identity to the registrantThe hardest gate, and the one that separates a stolen identity from a bad carrier. Authority verification and identity verification are different controls.
05 · lookup + judgementInsurance, both kindsPublic liability is on the federal file. Cargo coverage for general freight is not on any federal file — it has to come from the issuing source.
06 · lookup + judgementAffiliation and historyShared officers, addresses and filing agents against entities that were previously shut down. Legitimate small carriers share these too, which is why this prioritises review rather than deciding anything.
07 · judgementPayment setupContract, remit-to, and any factoring assignment — confirmed out of band, against details held before the request arrived.
08 · lookup + judgementDispatch and in transitThat the tractor, trailer and driver presenting are the ones the vetted carrier dispatched, and that the vetted carrier knows about this load at all.

The authorities

What each source proves — and what it does not.

Most vetting failures are not missing checks. They are checks that were run correctly and then read as proving something they never proved.

Source	What it proves	What it does not
USDOT number	An entity registered with the agency and was issued its unique identifier. It is the primary key, and FMCSA's new registration system keeps it as the single identifier.	Prove fitness, solvency, insurance adequacy — or that whoever is contacting you controls that number. MC and FF docket numbers are slated for eventual removal, but that was explicitly deferred, so they persist today.
Operating authority status	Whether authority is active, and its type. A party arranging transport for compensation needs broker authority; a motor carrier may not broker without registering as one.	Say anything when the authority is genuine and stolen. Every authority check passes in an identity-theft case, because the identity is real.
Company Snapshot and the public record	A public snapshot of registration, operating status, authority, fleet size, and inspection and crash counts.	Describe the current fleet, or guarantee the phone and email on file reach the real operator. Much of it is carrier self-reported, and self-reported data is the cheapest thing in the world to falsify.
Compliance categories (formerly BASICs)	The underlying inspection, violation and crash data is public for every carrier.	Give you a property carrier's percentiles or alerts — those were removed from public view under the FAST Act and are not available. Any product showing a property carrier's “CSA score” is showing public underlying data, its own proxy, or something reached under licence. Those are three different things.
Safety rating	That a compliance review happened, and what it found at that time. Satisfactory, Conditional or Unsatisfactory.	Mean the carrier is safe now. In the agency's own words a rating is a snapshot that “may be misconstrued as an FMCSA approval of the current operations.” And most carriers have no rating at all — 11,671 compliance reviews in FY2019 against a population above 567,000.
Insurance certificate	That a policy existed when the certificate was issued.	Create coverage, bind the insurer to you, or reveal exhausted aggregates, exclusions, deductibles or a cancellation already running. A policy can be cancelled on 35 days' notice, which is why this is monitored rather than sampled. A PDF certificate is also trivially editable — the structural answer is to obtain it from the issuing source, never from the counterparty.
Broker surety bond (BMC-84)	That financial security of $75,000 is on file, as the statute requires.	Protect the broker. It runs to shippers and motor carriers, for the broker's failure to carry out its agreements. It is an aggregate, not per-claim; on insolvency claimants share pro rata. One significant loss exhausts it, and it is not cargo insurance.
SCAC	That a code was issued to a named party and is current.	Say anything about operations or authority — in the issuing body's own framing it confirms identity, not operations. It is issued by a trade association and renewed annually, not a federal record.

The schemes

How the money actually leaves.

Each of these is written from the victim's side of the desk — how it runs, what it looks like while it is happening, and what the loss turns out to be. That is the useful view, because at the time every one of them looks like a normal booking.

Double-brokering

How it runs: A party accepts a load as the carrier, then re-tenders it to someone else and keeps the spread — frequently never paying the carrier that actually hauled it. Chains can run several layers deep.
What you see at the time: A normal booking that delivers normally. The load is often entirely fine. The problem surfaces weeks later as a payment demand from a carrier nobody has heard of, holding a signed proof of delivery, usually with a factoring company behind it.
What the loss is: Paying twice, an orphaned cargo claim against a party that was never underwritten for the load, and a record in which the broker cannot say who actually hauled the freight.

Carrier identity theft

How it runs: The attacker wears a real carrier's identity — registration numbers, legal name, sometimes a genuine insurance certificate — and substitutes its own phone, email and remit-to. The agency itself names account hijacking and the sale of carrier numbers among the patterns it is seeing.
What you see at the time: A carrier with a long clean history and real insurance on file. Every authority check passes, because the identity is real. The only false thing is who is on the other end of the phone.
What the loss is: A stolen load, or payment diverted away from the real carrier — which then presents its own invoice. The real carrier is a victim too, and often learns of it from the broker.

Fictitious pickup

How it runs: Wearing a stolen or freshly minted identity, the thief books a load and simply collects it with correct-looking paperwork. The freight is gone before anyone knows a theft happened.
What you see at the time: A clean booking and an on-time pickup. Then silence, or a plausible delay story, then nothing.
What the loss is: Total loss of the commodity, and a cargo claim with no solvent insured respondent — the carrier either does not exist or was never the entity named on the certificate.

Remit-to and banking change fraud

How it runs: A change of payment details arrives for a genuine carrier or factor, often from a real address inside a real email thread taken from a compromised mailbox. The freight moves perfectly; only the money goes astray.
What you see at the time: A routine administrative request with a plausible reason, frequently with a convincingly formatted release letter or assignment notice attached.
What the loss is: You pay the wrong party and still owe the right one. Recovery depends on same-day banking intervention and is usually partial. The highest ratio of loss to sophistication in the whole field — it needs no trucks and no cargo.

Phantom carriers

How it runs: A newly registered or never-operating entity with no equipment, no drivers and no intention of hauling, created to take loads, collect advances, or re-broker. The agency names fake initial registrations explicitly.
What you see at the time: Thin, but not disqualifying. Authority active, insurance filed, no adverse history — because there is no history at all. Absence of negatives reads as absence of risk, and that is the trap.
What the loss is: Theft, re-brokering, or an advance that goes nowhere. A clean record and no record are not the same thing.

Chameleon carriers

How it runs: An operator shut down or facing enforcement re-registers as a new entity, often reusing officers, addresses, phones or equipment, to shed a safety or payment history.
What you see at the time: A new entity with a clean slate — which is precisely the intended effect.
What the loss is: Elevated crash and claim exposure, and a selection decision made against a history that was discoverable. The applicable statute already requires an applicant to disclose common ownership, management or control with another carrier, forwarder or broker in the preceding three years.

Fraudulent factoring assignment

How it runs: A forged or stale assignment notice redirects receivables; or a double-broker factors an invoice for a load it did not haul; or two factors claim the same receivable after a carrier switched without a proper release.
What you see at the time: An assignment notice on letterhead, an invoice, and a valid proof of delivery. Factoring is ubiquitous and legitimate, and that volume of genuine traffic is the cover.
What the loss is: Paying the wrong assignee while remaining liable to the right one — and occasionally competing claims from two factors plus the carrier at once. Note that the innocent version, a stale notice with no release, produces identical mechanics.

Tracking manipulation

How it runs: Location telemetry treated as ground truth when its provenance cannot be established — a shared link, a third-party app, or a driver's phone, all of which are assertions by the counterparty.
What you see at the time: A tracking link that moves plausibly.
What the loss is: Delayed detection of a diversion, which is what turns a recoverable incident into a total one. Worth keeping in proportion: this matters as corroboration failure rather than as a scheme in its own right.

The controls

What each control actually stops.

Described as categories, deliberately. No single control stops a determined attacker, because the schemes attack different layers — authority, identity, documents, payment instructions and possession. What an attacker cannot easily defeat is a control that does not rely on information the attacker supplied.

Control	Where it sits	What it stops	What it does not
Identity verification	Capture, binding, and again at dispatch	Stolen identities, phantom entities, and the front end of a fictitious pickup.	Nothing about a real carrier that behaves badly.
Authority verification	Resolution and load suitability	Operating while revoked, and the flavour of double-brokering where a carrier has no broker authority at all.	Anything where the authority is genuine and stolen.
Insurance verification	Onboarding, then continuously	Uninsured-exposure surprises and a large class of document forgery, when the certificate is obtained from the issuer rather than the counterparty.	Mid-term cancellation and aggregate exhaustion — which is why the 35-day cancellation window makes this a monitoring problem, not a checklist item.
Out-of-band confirmation	Every payment-instruction change	Remit-to fraud and much assignment fraud — the two highest loss-per-effort schemes.	Nothing involving the freight itself. Its strength is structural rather than secret, which is why it is the one control worth advocating in public.
Equipment and driver corroboration	At pickup	Fictitious pickup at the last defensible moment, and surfaces unlawful re-brokering at the only point where it is directly observable.	Depends on shipper-side execution, which a broker does not control. Pretending otherwise is an overclaim.
Affiliation and pattern analysis	Onboarding and review	Chameleon entities and organised rings, which are invisible to any single-entity check.	The highest false-positive cost of anything here. Legitimate small carriers share addresses, agents and phones constantly, so this prioritises human review and decides nothing on its own.
In-transit verification	After dispatch	Nothing outright — it compresses time to detection, which is what decides whether a diversion is recoverable.	Only as good as the provenance of the telemetry. A link the counterparty controls is not independent corroboration.

Why the thresholds are not on this page

Freshness windows, scoring weights, which signal combinations raise a hold, which fields are checked and which are sampled, and what routes a load to enhanced review — none of it appears here. Published, it tells an attacker precisely what to forge and what to ignore. There is a second reason that points the same way: since the 2026 decision above, a broker's vetting record is litigated evidence, and anything a vendor publishes as a diligence standard becomes discoverable argument against its own customers. We go through the mechanics with you under NDA, not on a marketing page.

The legal frame

The rules that decide who pays.

Orientation, not advice — but these are the provisions that turn a vetting decision into a liability, and most of them are misquoted more often than they are read.

Vetting is now litigated evidence

A unanimous Supreme Court held in May 2026 that the FAAAA's safety exception saves a state-law negligent-hiring claim against a freight broker from preemption: requiring a broker to exercise ordinary care in selecting a carrier “concerns” motor vehicles. The pleaded basis for the broker's constructive knowledge was the carrier's Conditional safety rating. Carrier vetting stopped being only loss prevention that day.

Montgomery v. Caribe Transport II, LLC, No. 24-1238 (14 May 2026)

What that decision did not decide

It assumed without deciding that the claim would otherwise be preempted, and said nothing about the merits or about what diligence is sufficient. It was a bodily-injury case; whether it reaches cargo-loss and purely economic claims is unresolved. Anyone telling you the standard of care is now settled is selling something.

Slip op., n.2; remanded

The record a broker must already keep

A broker must keep a record of each transaction including the name, address and registration number of the originating motor carrier, and must retain it three years; every party to the transaction has the right to review it. The implication is pointed: a double-brokered load makes that record false, because the regulation presumes the broker knows who hauled the freight.

49 CFR § 371.3

Unregistered brokerage reaches individuals

A motor carrier may not broker transportation unless registered as a broker. Where a party provides brokerage without being registered and bonded, liability runs to the injured party for all valid claims without regard to amount — and jointly and severally to the entities involved and to their individual officers, directors and principals. The civil penalty is inflation-adjusted and currently $13,647 per knowing violation, not the $10,000 in the statute.

49 U.S.C. §§ 13902(a)(6), 14916; 49 CFR § 1022.4(b)

The bond is not the safety net people assume

Financial security of $75,000 is required regardless of the number of branch offices. It protects shippers and carriers from the broker, not the broker from anyone. It is an aggregate; on insolvency the surety advertises and claimants take a pro rata share if claims exceed it. New suspension machinery took effect in January 2026, tightening the timeline considerably.

49 U.S.C. § 13906(b)(3); 49 CFR § 387.307

Where cargo liability actually lands

Carmack attaches to the receiving, delivering, or line-haul carrier, and brokers are generally outside it — though an intermediary that holds itself out as a carrier risks being treated as one. In a fictitious-pickup loss that gives the shipper a right against a carrier that does not exist, so the real recovery route is the broker's own contract terms. The familiar nine months is a floor on what a carrier may impose, not a universal deadline.

49 U.S.C. § 14706(a), (e)(1)

Instrumentation

What a pilot reports back.

Metrics about the pilot's own behaviour, which is what earns trust in a first engagement — and which give away nothing about how anything is detected.

Decision latency, and what it gates

Time from carrier interest to verdict at median and 95th percentile, plus the share of loads where vetting was the step holding up the tender. Freight dies of latency, and a control that slows tender gets bypassed — bypass is the real failure mode.

Automated resolution, with a reason mix

Share closed with no human touch, and its complement broken down by why review was needed. The reason mix, not the headline rate, says whether it is working.

Source availability and data age

Per-source success and timeout rates, the age of the data each decision rested on, and fail-open versus fail-closed counts stated explicitly. A system that silently approves when a source is unreachable is worse than no system.

Hold outcomes both ways

Holds cleared on review as a false-positive proxy, against incidents on loads that cleared as an escape proxy — with the honest caveat that fraud surfaces late, so early escape rates flatter the system.

Re-vet coverage and time to detection

Share of active relationships re-verified within policy, lifecycle events caught, and the median time from event to detection. This is where a cancellation window becomes a measurable commitment instead of a claim.

Override and exception rate

By user and by reason. Controls rarely fail technically; they get overridden under time pressure. A pilot that does not measure this is not measuring the risk.

The engagement

Your onboarding flow, three weeks.

Fixed scope and a fixed price: $4,500. We take your onboarding flow as it runs today, wire up the deterministic verification, put the judgement steps into a queue with the evidence attached, and instrument the whole thing from the first day — so what you get back is your own exception profile rather than our opinion of it.

What we need: your carrier agreement and onboarding packet, your current steps however informal, a sample of recent onboardings including any you later regretted, and one person in operations or risk who answers questions.

Book a 30-minute call with our expert

Delivered

Authority and identity verification wired into onboarding
Insurance monitored against lifecycle events, not sampled
Out-of-band confirmation on every payment-instruction change
Review queue with the evidence attached and a decision of record
The vetting record itself, in the shape a dispute will ask for it

FAQ

Questions a risk director actually asks.

Can you just check FMCSA and tell us a carrier is safe?

No, and the agency is clearer about this than most vendors. A safety rating is a snapshot from the time of an investigation, and in FMCSA's own words may be misconstrued as approval of current operations. Most carriers have no rating at all — there were 11,671 compliance reviews in FY2019 against a population above 567,000. A clean record and no record look identical in a database and mean very different things.

Can you verify a carrier's cargo insurance from a federal database?

Not for general freight — and this surprises people. The federal filings cover public liability, where the floor for general freight is $750,000, not the $1,000,000 that circulates. The only cargo minimums in the regulation attach to household goods. The widely quoted $100,000 cargo requirement is a broker contractual convention, not a federal rule. Cargo coverage has to come from the issuing source and be read for its terms, not just its limit.

What actually stops double-brokering?

No single control, because the schemes attack different layers — authority, identity, documents, payment instructions, and possession. What an attacker cannot easily defeat is a control that does not rely on information the attacker supplied: identity bound to the federal record through channels derived from that record, certificates obtained from the issuer, payment changes confirmed out of band, and corroboration of who actually showed up at the dock.

Why doesn't this page list your detection rules?

Two reasons, and they point the same way. Fraudsters read vendor pages, so publishing freshness windows, scoring, which signals trigger a hold or which fields are checked would tell an attacker exactly what to forge and what to ignore. And since the Supreme Court's May 2026 decision, a broker's vetting record is litigated evidence — anything a vendor publishes as a diligence standard becomes discoverable argument against its own customers. We will walk through the mechanics with you under NDA.

Does the broker bond cover us if a carrier defrauds us?

No. The $75,000 financial security protects shippers and motor carriers from the broker — it does not protect the broker, it is an aggregate rather than per-claim, and on insolvency claimants share pro rata. It is also not cargo insurance. This is the most common misconception in the whole domain.

Isn't FMCSA fixing this with identity proofing?

It is moving in that direction, and the direction is real — identity proofing for new applicants and verification of existing registrants as they first reach the new system. But it covers registration, not the moment a stranger offers to haul your load, and parts of the programme were explicitly deferred. Our work is the private-sector mirror of a federal direction of travel, not a substitute for it.

What does a pilot cost?

A pilot is $4,500 fixed: three weeks, your onboarding flow, the deterministic verification wired up and instrumented from the first day, with the exception queue and the review reasons reporting themselves. The scope is agreed in writing before anything starts.

What do you need from us to begin?

Your carrier agreement and onboarding packet, your current vetting steps however informal, a sample of recent onboardings including any you later regretted, and one person in operations or risk who can answer questions.

Let's build it

Book a 30-minute call with our expert

Bring your onboarding packet and one onboarding you regretted. You will leave the call knowing where your process is load-bearing, where it only looks like it is, and what a pilot would cover — $4,500 fixed, three weeks.

Book a discovery call

Reply within 1 business day · India & USA

Statutory and regulatory references are to the United States Code and the Code of Federal Regulations as cited, and to the published decision named, and are given for orientation only — this page is not legal advice and no part of it describes a standard of care. Rulemaking and case law move; verify the current position before relying on any of it. Figures attributed to the agency are from its published notices. No customer data, configuration or detection logic appears here. We do not make safety fitness determinations, which are a function of the regulator.

Also in this blueprint: freight EDI onboarding and repair, invoice and accessorial audit, document intake, exception and ETA agents, and the rest of the AI TMS Blueprint.