The August 1, 2025 forum: flower labels fail a statewide test, and the surveillance program takes shape

Colorado's Science & Policy Forum met August 1, 2025, five months before the new testing rules took effect. We transcribed the Division's published recording (about 96 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. The published recording begins mid-session: the meeting was disrupted and restarted in a locked-down webinar format, with the co-chairs apologizing on the record for "grossly offensive and inappropriate language and images" at the start and moving all participation to moderated Q&A [0:00:00, 1:30:52]. What the recording preserves is the substance. And the substance was two of the biggest threads of the coming year.

Notes from the editor

First, something the public should sit with: the conditions the state asks its own scientists to work in. The people running the state's surveillance tests are highly trained chemists and technicians, and Colorado houses them in a 30-year-old building whose air handlers failed one after another over the winter (one exploded) and where metals testing went down entirely for lack of exhaust ventilation. In a chemistry lab, that ventilation is what stands between the people working there and the solvents and toxins they handle. And this is not an abstract facilities story: what the building's failures have delayed is health and safety testing. Every product on a shelf passed a lab test before it was packaged. But mold grows between packaging and sale, mites and pests can get in, and a mishandled or badly made package can let in anything. Shelf surveillance is the only testing that looks at a product the way a customer actually receives it. The Division defended the program on exactly that ground at this meeting. Is post-packaging contamination a big health problem? Probably not. But nobody knows, because the checks that would tell us haven't run. And things you don't know can hurt you.

Keep the scale in mind, too. This testing system consumes real money and many hours of very talented scientists' and technicians' time. Spending that efficiently matters. And to their credit, the regulators in these recordings are visibly trying to. But remember who actually pays for all of it: you. The cost of every required test lands in the shelf price, and the state collects its cannabis taxes on top. People who hand the state that money deserve a state that provides adequate scientific facilities in return. That is part of what makes it safe to give the state your money. When flower surveillance slips a year, the easy move is to blame the lab. Blame the building. A state that wants trustworthy cannabis testing has to fund the facility, not just the mandate.

Now, here's the situation this meeting put on the record. An independent researcher (the holder of Colorado's only marijuana research-and-development license) bought 281 products off dispensary shelves across the state and had them tested against their own labels. The study was run blinded: the lab tested each sample without knowing what its label claimed, while a University of Colorado team held the key matching samples to products. That design matters: a lab that can't see the label can't be tempted, even unconsciously, to nudge a number toward it. The study was presented to the state's own science forum, in public, in August 2025. No one at the meeting, regulator or industry, disputed its findings: that concentrate labels were honest about 96% of the time, and that flower labels were wrong more than half the time, mostly claiming more THC than was in the jar.

The question a reader should ask first about this meeting is the one I asked: did the state change how flower is tested as a result of these findings? The answer, a year later, is no. The January 5, 2026 rules left the strain-based flower potency schedule (the structural cause the Division's own co-chair identified at this meeting) in place, and the state's own check of flower labels has still never run.

The study's quieter finding aged into the bigger story. CBG (cannabigerol) is the plant's starter cannabinoid, the raw material it converts into THC and CBD as it matures. It's biologically active and barely studied in humans. The study found it in nearly every concentrate, sometimes near 4%: a real dose of a compound that wasn't on the label. The concern isn't that CBG is known to be harmful; it's that people were taking it without being told, in amounts nobody was tracking. That unlabeled-minor-cannabinoid sighting was an early glimpse of what the Division's science director confirmed on the record in June 2026: hemp-derived cannabinoids, intentionally added, now riding at ratios that rival THC. Be careful to split that in two. Adding hemp-derived cannabinoids to a product is arguably a legitimate formulation choice, provided the hemp gets safety testing, which the January 5 rules now require for any hemp entering the regulated market. Converting hemp CBD into intoxicating cannabinoids with solvent chemistry is a different animal entirely: uncontrolled synthesis that leaves behind byproducts the standard test panel was never designed to see. The state agrees: it wrote a dedicated conversion-solvent screen into the January 5 rules. That screen has been law for seven months and has never run, because no lab in the state is certified to perform it. Until it runs, the legitimate path and the bad one walk through the same door, indistinguishable.

Since this meeting

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

Resolved. The surveillance program presented here as a draft became real: its design was locked at the September 2025 meeting, it launched on schedule, and its edibles phase was executed and publicly reported in June 2026: every safety test passed; only 47% of samples matched their label on both potency and homogeneity.

Still in progress. The label-wording chaos this study documented (THC, total THC, THCA; even the academic team had to guess) is finally the target of rulemaking: standardized reporting units and labeled-cannabinoid testing requirements were previewed in July 2026 for the August rulemaking season. Proposals, not yet law, and driven by the edibles surveillance data, because that's the only data the state has produced.

Put off or neglected. Flower, the headline of this meeting. The category this study said fails more than half the time is the one category the state has never retested: flower surveillance slipped from early 2026 to fall 2026 behind test-method validation and physical breakdowns at the state lab's 30-year-old building: its air-handling units failed, one explosively, taking the safety ventilation a chemistry lab can't run without offline for weeks. No rule change has touched how flower potency is tested or labeled. And where the state does have retest data (edibles), the question this study opened was put to the forum directly at the August 7, 2026 meeting and answered with "I don't think we have an answer to that." The who-pays-for-samples question posed here was never settled. And the presenter's 30%-THC skepticism was not returned to in any later recording we transcribed.

What was brought up

A statewide test of whether labels tell the truth [0:05:04]. The holder of Colorado's only marijuana research-and-development license (a researcher with an ISO-certified analytical lab, working blinded with a University of Colorado Boulder team) presented a published study that purchased 281 products from 52 recreational dispensaries across the state and compared the label to what the lab measured. The headline findings as presented: concentrates were honest (about 96% matched their label within a ±15% variance) while flower failed more than half the time, with roughly 56–57% of flower products outside the variance band, most of them labeled higher than they measured, some lower, with a wide spread in both directions [0:10:08]. Secondary findings: CBG turned up in nearly all concentrates, sometimes near 4% (an unlabeled, arguably meaningful dose of a little-studied cannabinoid) [0:50:30], and label wording was chaotic enough (THC, total THC, THCA) that even the study's own academic team had to guess what some labels claimed [0:20:11].

Why flower fails and concentrates don't [0:25:14]. The Division's co-chair connected the result to the rules themselves: every concentrate batch must be potency-tested, while flower potency rides on a strain-based schedule (the first four harvests, then quarterly), so a label can describe a test months removed from the jar it's on. The presenter added the physics: concentrates are homogeneous by manufacture; flower is a living plant that varies plant to plant and dries season to season.

The surveillance program's working design [1:05:35]. The Division walked through its draft shelf-surveillance program: roughly 150 samples over three months, one product category per month (edibles, then concentrates and infused pre-rolls, then flower and pre-rolls), producers selected by a market-share-weighted stratified draw, samples collected from store shelves by MED investigators "prepared just like they were serving a customer" [1:20:46, 1:25:50]. A staff physical scientist presented the collection SOPs: sanitized coolers, temperature and humidity data loggers, evidence-bag chain of custody, same-day courier delivery to the reference lab where possible [1:20:46].

What was nailed down

• The study's design points, as presented: 281 products, 52 stores, purchases spread statewide under daily limits, whole-sample homogenization including stems, blinded analysis with the academic team holding the key, results published in a peer-reviewed journal [0:05:04–0:20:11].
• The surveillance program's structure: category-per-month sequencing, stratified market-share sampling, store-level substitution rules, and draft collection SOPs shared for stakeholder review, with a September launch target [1:30:52].
• Health-and-safety advisories stay case-by-case: a failed surveillance test triggers the existing MED–CDPHE consultation process, not an automatic advisory [1:05:35].
• The format change: after the disruption, the forum moved to webinar mode with moderated Q&A, "to control access" [0:00:00].

What was left open

• Whether surveillance should duplicate what the study already showed. The Division openly asked whether concentrate potency testing should be dropped from the surveillance program, since the published data showed concentrate labels largely honest: a resource-versus-redundancy question left for the September meeting [1:05:35].
• Who pays for the samples. Surveillance samples are taken from store shelves without payment; whether cultivators and manufacturers would backfill stores' losses was posed as a question licensees should settle among themselves. Unresolved [1:10:38].
• Where sampling should happen. A commenter argued samples should come from grows and manufacturers before market entry, not from sales floors; the Division defended shelf sampling as the only way to see what actually reaches consumers, including transport and storage effects [1:15:44].
• How to fix flower labeling. The presenter's view: tighten label verbiage rules and testing-lab practices before rebuilding the whole system: "there are tweaks that you can make... without blowing up the entire system" [0:55:31]. The underlying question (whether flower potency labeling can be accurate at all under strain-based testing) was named but not answered.
• What 30%+ THC flower labels mean. The presenter's skepticism, as a biologist: chlorophyll is the plant's most abundant compound at around 30%: "to have this many products testing that high begs the question of how we're getting there" [0:40:23].

This summary was drafted by our AI desk from a machine transcript and has been reviewed and approved by our editor (2026-08-12). The research findings described are the presenter's published claims as stated at the meeting. This site has not independently reviewed that study. The Division's published recording is the source of record.