Federal investigators visited a registered commercial driving school and asked where classroom instruction took place. The answer, as the agency recorded it, was that the classroom was inside a school bus, in the back of a trailer.
That school was on the federal Training Provider Registry. It had been there legally, because of how a provider gets on the list.
Nobody goes and looks first.
What a CDL mill is
There is no statutory definition. The clearest official description of the mechanism comes from a federal indictment in Louisiana, which maps the fraud onto the three steps a commercial driver has to clear.
Two motor vehicle office employees allegedly completed written knowledge tests for applicants using internet searches. Two operators of truck driver training businesses allegedly reported training completion falsely in the federal database and submitted false passing scores. A skills test examiner allegedly created fraudulent score sheets.
Knowledge test, training record, skills test. All three, all bought. Three defendants pleaded guilty in July 2026 to bribery concerning programmes receiving federal funds, and prosecutors say more than 100 people obtained commercial licences through the scheme.
That is a CDL mill: not a bad school, but a functioning parallel process for producing the paperwork of a qualified driver without producing a qualified driver.
Getting on the federal list is a form, not an inspection
The Entry-Level Driver Training rule took effect on 7 February 2022 and requires new commercial drivers to complete training from a provider listed on the federal registry.
Read how a provider gets listed. Under 49 CFR 380.703, the provider must follow the curriculum, use compliant facilities and vehicles, employ qualified instructors, be authorised under state law, and agree to allow audits. It then submits a registration form electronically, and the agency issues it a registry number.
Every one of those requirements is self-attested. There is no pre-listing inspection anywhere in the subpart. The audit authority exists, but it is entirely after the fact.
The registry is therefore a database of claims. Today it lists 30,992 active locations.
The rule removed the one number that would make a school auditable
Here is the design decision that makes the rest of it hard to police, and it is in the rule’s own text.
From the final rule: “There is no minimum number of hours that driver-trainees must spend on the BTW elements of the core or endorsement curricula.” Behind the wheel, no minimum. Instead, the provider must not issue the certificate “unless the driver-trainee demonstrates proficiency in performing all required BTW skills.”
For a serious school that is defensible teaching. Competence, not clock-watching.
For a mill it is an unfalsifiable claim. An investigator cannot contradict a seat-time record that federal law never required anyone to keep. Look at what the July 2026 investigations actually found: no space on the range to complete manoeuvres, an instructor without the correct licence, no records of any assessments kept, and the classroom in the trailer. Those are facilities and records failures, because facilities and records are the only things with a standard attached.
The quality of the training itself has no federal metric to fail.
How the government finds the bad ones
This is the sentence that tells you the most about the architecture.
Describing an emergency removal of more than 110 training providers, the agency explained its method: it reviewed roadside inspection records for drivers cited over English language proficiency, then compared those drivers against the registry to identify providers that had repeatedly certified drivers who did not meet federal qualification standards.
Read that again as a process. The federal government identifies a fraudulent school by waiting for its graduates to be pulled over.
It works, and the agency deserves credit for thinking of it. But it is pure downstream detection. With no verification at the front end, the quality control instrument for a 30,992-location registry is a state trooper at a weigh station, and the failure is only discovered once an unqualified driver is already behind eighty thousand pounds on a public road.
We covered the first big sweep when the feds shut down 110 truck driving schools overnight. In July 2026 the agency sent 175 investigators to 40 states to conduct nearly 400 investigations, producing more than 160 proposed removals. That is a real enforcement effort. It is also roughly one percent of the registry. At that rate, visiting every listed location once would take the better part of a century.
Be careful with the headline numbers
The counts in circulation do not reconcile, and anyone writing about this should say so.
One departmental statement in July 2026 cited more than 9,500 unqualified training schools purged. A statement in August 2026 cited more than 8,000 — a smaller number, later. The live registry today shows 9,958 locations removed.
Three figures, three sources, and they cannot all be counting the same thing.
There is also a strong reason to think most removals are not fraud at all. The registry itself notes that the agency contacts providers whose locations have submitted no training certifications in the past twelve months. A school that quietly went out of business and a school that faked its records both end up as removals, and the published totals do not separate them.
Treat any “thousands of fraudulent schools shut down” claim with care until the breakdown is published.
Two federal systems, and each assumes the other is watching
The structural weakness is a seam. Training is federal, administered through the registry. Testing is a state function, and states routinely delegate skills testing to third-party examiners.
The Louisiana case ran straight through that seam — state motor vehicle clerks on one side, federally registered training operators on the other, a third-party examiner at the end. Three control points, three different custodians, none of whom answered to the same supervisor.
And the federal requirement for states to police their examiners is thinner than you would expect. 49 CFR 383.75 requires a state to inspect each third-party tester on site at least once every two years, and at least every two years to either covertly take tests as if it were an applicant, co-score, or re-test a sample of drivers.
A biennial secret shopper is the entire structural defence against a corrupt examiner.
What federal leverage over states looks like in practice
When the federal government does move on a state, the mechanism is money, and New York is the complete worked example.
An audit of 200 sampled non-domiciled commercial licence records found 107 issued in violation of federal law. Findings were announced in December 2025. The state responded in March 2026. On 16 April 2026 the department made a final determination of substantial non-compliance and withheld $73,502,543 — four percent of two federal highway funding programmes.
The penalty structure under 49 CFR 384.401 rises to eight percent in a second year. The nuclear option, decertification under 384.405, would bar a state from issuing commercial licences at all, and has not been used.
It is worth asking whether withholding road construction money is a well-aimed instrument for changing how a licensing office processes paperwork. The people who lose the highway funding are not the people who issued the licences.
The oversight of the overseer has not reported yet
One more gap, and it is the one that should make everyone cautious about the numbers above.
On 19 September 2025 the Department of Transportation’s Inspector General opened an audit of the agency’s oversight of state commercial licence knowledge and skills testing programmes. It was self-initiated, not requested by Congress. It is still open.
There appears to be no completed Inspector General audit of the Training Provider Registry at all.
Which means every enforcement figure in public circulation is agency self-reported, arrives by press release rather than report, and — as the inconsistent school counts show — is not internally consistent. The independent audit that would reconcile them is pending.
One more thing worth separating
The enforcement campaign has been presented largely through the frame of foreign drivers and immigration enforcement, with joint announcements across departments.
The prosecutions in the record look different. A Massachusetts state trooper was sentenced for giving passing scores to applicants marked “golden” — three who failed and five who never took the test — with free bottled water among the inducements. A Louisiana restaurant owner allegedly organised payments to motor vehicle office employees.
Those are domestic corruption cases about public officials selling credentials — a distinction we drew when Washington pulled 24,000 truckers off the road. They may be occurring alongside the other problem, but they are not the same problem, and conflating them makes both harder to fix.
What to remember
Forget the removal totals. Remember what the registry actually is.
A federal list that tells a prospective driver a school is approved, built entirely on that school’s own word, with no requirement that anyone visit before the listing goes live, no minimum hours that could be checked against a timesheet, and a discovery process that begins when a graduate gets stopped at the roadside.
The truck driving school with a classroom in the back of a trailer was not hiding from the federal government. It was on the list.
Should the federal government actually verify these schools instead of taking their word for it? Share your take in the comments.