The June 12 forum: third-party sampling dies in the room, and the edibles results finally land (clean product, unreliable labels)

Colorado's Science & Policy Forum met June 12, 2026 (delayed twenty minutes by a Zoom-account failure that forced a new meeting link). We transcribed the Division's published recording (about 1 hour 55 minutes) with a speech-recognition model and drafted this summary from that transcript; our editor has reviewed and approved this page, and the recording remains the source of record. Two threads that ran all year reached their destinations here: the third-party sampling proposal met the room and lost it, and the surveillance program's edibles results, promised since November, were finally presented in full.

Notes from the editor

Two sentences from this meeting matter more than everything else. The regulator's: businesses are testing "at the lab that passes their product the most frequently instead of developing partnerships": lab-shopping, named by the state, on the record. And the data's: only 47% of edibles matched their own label on both potency and homogeneity, while every safety test passed. Clean product, unreliable labels: the year's verdict, delivered in one meeting.

Credit the source of both: the regulator volunteered them. Nobody forced the Division's science director to name lab-shopping, flag the reduced-testing-allowance problem, or read the label failures into the record. That is an agency disclosing findings against its own program's comfort, and it's exactly the institutional behavior a public forum exists to produce.

On the minor cannabinoids: hemp-derived cannabinoids intentionally added at ratios rivaling THC splits into a legitimate practice and a dangerous one, and the state currently can't tell them apart. I've laid that argument out in my note on the August 1, 2025 page; this meeting is where the underlying finding landed. The short version: added hemp isolate with safety testing is arguably a fair formulation choice. And the January rules now require that testing. Solvent-converted hemp is another animal entirely, and the screen that would catch it has been law since January without a single certified lab to run it.

And the sampling decline: reasonable people can decline a costly rule. The geography and the economics are real. But the room that declined it had spent months agreeing that adulteration is real and that enforcement is the variable that matters. Declining the rule obligates you to the alternative. The alternative hasn't appeared.

Here's what I'd like to see happen next, and none of it needs a new rule. A producer offered back in November to put its own certificates head-to-head against the state's results on the same batches. Take them up on it. And run the cheap arm nobody's blocked from running: shelf product sent to the private labs, which can legally receive anything (it's the state lab that's federally constrained). This doesn't need to be a peer-review-grade double-blind study. It needs to be a production-level experiment. If the private labs agree with each other on crossed samples, then the noise isn't between the labs, and the next cheap experiment aims at sampling or the state-lab comparison. If they don't agree, we've found it. Small experiments, run in sequence, each one telling you where to point the next. That is how a working system debugs itself, and every piece of it is available today.

Since this meeting

Where the threads this meeting opened stand as of August 2026, from the recordings that followed.

Resolved. The analyte submission form presented here went live in July. The full phase-one report was published and linked in the forum's public folder. July's meeting was moved to the 10th and happened.

Still in progress. Everything the data pointed at is now rulemaking-in-motion for the August season, per July's preview: standardized reporting units, a defined homogeneity basis, potency and homogeneity results traveling together, labeled-cannabinoid testing, and reduced-testing-allowance enforcement. The metals and Aspergillus technical breakout was planned for later in the summer.

Put off or neglected. Which labs the failures clustered at was asked directly and not disclosed. The homogeneity spread data (percent RSD) a manufacturer requested was not provided in the recordings we transcribed. And the enforcement alternative the sampling opponents championed (minimum fines, license revocation for adulteration) followed the pattern of the whole year: endorsed again, proposed never.

What was brought up

Third-party sampling, fully built and then declined [0:05:06]. Two Division scientists presented the complete framework: every rule that would change (collector employed by testing facilities, in-person observation by the sampled licensee, an expanded per-batch collection record, and a new verification rule: physically weigh or count at least one production batch against the inventory system, ±10%, or stop and notify the Division) [0:15:11–0:20:12], the information-gathering results (cost estimates "anywhere from a nominal increase... up to 300%," timing of at least an hour per visit plus drive time, likely reduction in the number of testing facilities) [0:10:08], and the honest closing question: is this the most efficient approach to the desired outcomes? [0:25:14].

The room answered. Operator after operator opposed it: the costs (one manufacturer, notably revising his own earlier tripling estimate after more research, now projected $150–200 per pickup, roughly $40–50,000 a year for a daily-pickup operation [0:35:16]); the geography (seven licensed labs, nearly all on the Front Range, one in Pueblo [0:30:15]); the economics of an industry that "has lost a billion dollars of revenue over the last four years" with a demand for professional economic analysis before any such rule [0:40:17]; and the alternative program repeated like a chorus: expand surveillance, inspect annually, set minimum fines, revoke licenses for adulteration [0:35:16, 0:40:17]. The lab side dissented: a laboratory representative supported the change as closing the custody gap ("we just test what we receive... clients don't really know what happens to a sample after they drop it off"), estimating costs at 5–10% [0:50:19]. The co-chair's conclusion, on the record: "there's likely not a consensus to move this as a science and policy recommendation", while noting nothing precludes the agencies from proposing rules on their own [1:05:42].

The edibles results, at last [1:20:56]. The state lab presented phase one in full. Forty-five edible samples (gummies and fruit chews dominating, plus powders, beverages, pills, and singles of other types) across three market-share strata. The safety panel was spotless: every microbial test passed (no total yeast and mold above the detection level, no salmonella, no STEC) and every elemental-impurities test passed (nothing quantifiable). The label panel was not: 35 of 45 passed homogeneity, 23 of 45 passed potency, and only 21 of 45 (47%) passed both [1:26:05]. Most potency failures were marginal (a manufacturer noted many missed by fractions of a milligram), a couple were large, and the most-failed analytes were delta-9 THC and, among products that labeled it, THCV [1:26:05].

What the regulator read in the data [1:31:08]. The Division's science director offered the year's most consequential analysis: roughly one-third of the failing samples appeared to involve reduced-testing-allowance problems: products riding a testing exemption that may not have been compliantly earned [1:46:14]; minor cannabinoids, often hemp-derived and intentionally added, now appear at ratios rivaling THC with labeling rules that haven't caught up; and (said plainly by the regulator, on the record) businesses are "deciding to test at the lab that passes their product the most frequently instead of developing partnerships" [1:36:09]. The state lab's own caveat cut both ways: the three evaluation metrics it used are not written in rule, so part of the label failure rate may be a measurement-definition mismatch between the reference lab and industry practice [1:31:08].

The analyte form shipped [1:05:42]. CDPHE presented the final survey-format submission form: six sections, facts and sources only, no submitter self-scoring, incomplete submissions not reviewed, with the system-impact analysis deliberately moved downstream to the agencies. The original stakeholder author blessed it, and the form heads to the Division's website [1:15:56].

What was nailed down

• No forum recommendation for third-party sampling: the year's loudest proposal ends 2026 without consensus [1:05:42].
• The edibles numbers themselves: 45 samples; 100% pass on microbials and metals; 78% pass on homogeneity; 51% pass on potency; 47% pass on both [1:20:56, 1:26:05].
• The full report's status: final publication approval in progress, with the summary already in the forum's public folder; discussion to continue at the July meeting [1:46:14, 1:51:15].
• A second technical breakout planned for metals and Aspergillus methods, later in the summer [1:51:15].
• July's meeting moved to the 10th (the first Friday fell on the holiday weekend) [1:51:15].

What was left open

• What the label failures mean. Lab inconsistency, manufacturing variance, out-of-rule metrics, or reduced-testing-allowance abuse: the data can't yet apportion blame, and the Division said the nuanced investigation is ongoing [1:31:08, 1:46:14]. A manufacturer asked for the homogeneity spread data (percent RSD) the presentation didn't show [1:41:11].
• Which labs diverged. Asked directly whether failures clustered at particular licensed labs, the agencies didn't disclose; asked how CDPHE knows its own results are right, the answer was ISO 17025 accreditation and proficiency-test participation [1:46:14].
• Whether one study justifies action: a standards-body participant cautioned against acting on a single 45-sample study; the labeled-cannabinoid-threshold idea (mandate labeling above a set ratio) drew support as a low-cost fix [1:41:11, 1:46:14].
• The rules the data points at: dose-uniformity-focused potency and homogeneity rules, minor-cannabinoid labeling, and reduced-testing-allowance enforcement. All now queued as rulemaking conversations, none yet proposed [1:31:08].

This summary was drafted by our AI desk from a machine transcript and has been reviewed and approved by our editor (2026-08-13). The Division's published recording is the source of record.