August 28, 2026
Ask It Twice
There is a cheap test for whether you are looking at a rule or at a summary of one, and it takes about a minute.
Ask the same question twice, worded differently. Then put the two answers side by side and look only at the numbers.
A rule holds still. It was adopted on a date, filed under a number, and printed in a code section, and it says the same thing at nine in the morning as it does at four in the afternoon no matter how you phrase the question. That immovability is not a nice property of rules. It is most of what a rule is.
A summary does not hold still, and this week it moved far enough to watch.
What moved
We asked what a Texas retailer faces for non-compliance. We asked it four ways over one morning.
The first phrasing returned fines up to $25,000 per violation, plus criminal referral.
The second returned up to $10,000 per violation, plus loss of the consumable hemp licence.
The third returned $10,000 again.
Then we opened the commercial page that the answers themselves credited. It puts a first offence at a Class A misdemeanour — up to a year and a $4,000 fine.
Three different numbers for one penalty, across four asks, in a single morning. Not three different rules. One question, asked four times.
You cannot plan against that. More to the point: you cannot tell, from inside any one of those answers, that the others exist. Each arrives complete, confident, and alone.
The thing that should have decayed and did not
[We wrote yesterday](https://gethempdash.com/learn/a-deadline-nobody-published) about a specific compliance timeline circulating among Texas operators — inspections commencing August 1, a thirty-day grace period, penalty-free surrender of banned inventory through August 31, roughly 2,400 registered retailers to be inspected between August and October, sales records audited back to June 1, criminal penalties only for willful violations. We went looking for the regulator that published it and could not find one, and the page it was credited to said the opposite of what it was cited for.
We re-ran it this morning, expecting it to thin out.
It did the reverse. It came back on a third, differently-shaped phrasing. And it had grown: out-of-compliance product would now be "detained and referred to law enforcement," and retailers should "retain receipts through December 2026 to demonstrate compliance during audits." Neither of those was in yesterday's capture. They are new, they are specific, they are operationally actionable, and they are credited to nothing.
We re-checked the agency's surfaces this morning rather than trusting yesterday's check. The Consumable Hemp Program page carries none of it; its only enforcement sentence is that businesses found in violation may have their licence or registration revoked. The Consumable Hemp Products FAQ carries none of it. The news-alerts index still carries no consumable hemp item at all for July or August 2026.
An unattributed claim that accretes detail overnight is not converging on the truth. It is doing the opposite of what a record does. A record gets more specific only when something is filed; you can point at the filing.
The same failure, running backwards
The same morning, a sweep for current Texas enforcement returned a police-and-DEA operation on nine shops — nine warrants, five arrests, second-degree felony charges, undercover buys including minors, product tested between 7% and 78% THC — placed in "late August 2026."
It happened in August 2024, in Allen, and we established that from three independent outlets that covered it at the time.
Look at what the summary got right. The shop count. The arrest count. The charge. The THC range. The day and the month of the arrest date. Every checkable particular survived intact. The only thing that moved was the year, and it moved by two.
That is the mirror image of the invented timeline, and it is worth putting the two next to each other. In one case, a rule that does not appear to exist arrives with a full set of dates. In the other, an event that certainly did happen arrives on the wrong date entirely — while keeping enough correct detail to pass every plausibility check you might run on it.
The shared property is not falsehood. It is specificity without a source. Both are precise. Neither is anchored.
What we are not telling you
We are not telling you there is no grace period. We did not find one, which is a narrower sentence, and an absence of an announcement is not an announcement of absence.
We are also not telling you the opposite — that the absence means inspections are imminent, or aggressive, or that anyone is being pursued. That is the inference this sort of correction habitually smuggles out the back, and it is exactly as unsupported as the thing being corrected. A debunked leniency claim is not evidence of severity.
Our record holds the enforcement posture on the Texas total delta-9 THC threshold as under review and asserts it in neither direction. Nothing this morning moved that field. A claim with no author cannot move it, whichever way it happens to point. [A raid is not a rule change](https://gethempdash.com/learn/a-raid-is-not-a-rule-change), and neither is a summary.
Worth noting as a control, because it cuts the other way: a named law firm published an advisory on this exact question on August 4, under its own byline, and supplied none of the missing numbers. No grace period, no inspection target, no audit lookback, no dollar figure. Its enforcement paragraph runs to one sentence about civil tools and licence revocation, and it advises businesses not to infer an absence of enforcement from the state of their inventory.
The pattern is consistent and it is not an accident: the closer a source sits to accountability, the fewer specifics it volunteers. When the unattributed answer is more detailed than the signed one, the detail is the warning.
The numbers that did hold still
We ran the same test on our own record, which is the only fair way to publish it.
The certificate requirement. 25 TAC §300.301(d) names twelve fields as a condition of sale — among them laboratory identification and contact information, sample and lot identification, analytical methods and instrumentation with limits of detection and quantitation, an expiration date, measurement-of-uncertainty parameters, a QR code verifying the authenticity of testing at an accredited laboratory, and results reporting three values: delta-9 THC, total delta-9 THC, and total THC per container. A certificate missing any of the three values is non-passing. Twelve, every time you ask. The field set is walked through in [The Twelve Fields a Texas COA Must Carry](https://gethempdash.com/learn/texas-coa-requirements-twelve-fields).
The registration figure, and the other registration figure. 25 TAC §300.502(g) sets retail registration at $5,000 per retail location, and again at renewal, with a $1,000 delinquency fee and no proration. Separately, the agency publishes what you actually pay: $5,150 per location for a one-year retail registration, stated to include the Texas Online fee, and $10,300 for a one-year consumable hemp product licence.
Those are two different quantities and both are stable. The rule's number is the rule's number; the amount payable is the amount payable. Neither wobbles when you rephrase the question, and the gap between them is a disclosed surcharge rather than a discrepancy — which is precisely how you can tell it apart from what happened to that $25,000.
The arithmetic. The threshold binds total delta-9 THC, inclusive at 0.3% or less, with a measurement-of-uncertainty allowance. Total delta-9 THC is delta-9 plus 0.877 times THCA. The multiplier is a molecular-weight ratio. It has never returned a different value to anyone.
Which records are settled, which are pending, and which are open in one direction only is set out record by record in [Which Texas Hemp Rules Are Unsettled, and Which Are Not](https://gethempdash.com/learn/which-texas-hemp-rules-are-unsettled).
And in the future tense, where it belongs: a federal provision, P.L. 119-37, is scheduled to take effect on November 12, 2026, and will change how the federal definition is measured. It has not done so. Any account describing it as currently in force is describing a date that has not arrived.
The habit
Ask it twice.
If the number holds, you have probably found the rule, and you should still go and read it. If the number moves, you have found a summary — and the question of what the rule actually says is still entirely open, no matter how confident either answer sounded.
The failure mode this catches is not a source being wrong. It is a source not existing. Reputation checks miss that. Plausibility checks miss it badly, because the specificity that makes a claim feel verified is exactly what gets generated when there is nothing to verify against.
Two asks. One minute. Do it before the number goes on a whiteboard.
Source: 25 TAC §300.301(d); §300.502(g); §300.301(a)(4)/(b)(4) and §300.302(b)(1)/(c)(3). Adopted rules filed with the Secretary of State March 2, 2026 (TRD-202601066), in effect March 31, 2026. Registration amounts payable as published by the Department of State Health Services. Enforcement posture: under review, asserted in neither direction.
The full record set, with each status shown rather than summarised, is at [the Texas Rule Center](https://gethempdash.com/texas-rules).
Want this handled for your store? See it working in one demo.
Book a Demo