The September 12, 2025 forum: the surveillance program locks its design, and the fight over transparency begins

Colorado's Science & Policy Forum met September 12, 2025, with one job: finalize the shelf-surveillance testing program for launch. 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 design work got done. The meeting's most heated minutes were about something the design didn't cover: how fast the public gets to see the results.

Notes from the editor

First, a disclosure: I attended this meeting myself, as a signed-up forum participant. For the avoidance of any doubt, I am not the "transparency advocate" described on this page, nor any other speaker quoted or characterized here. I was in the audience. Where my memory of the room informs these notes, it stays in this section; the meeting-record sections above rest on the Division's published recording alone.

The fight in this meeting was over how fast the public learns what the state finds, and you should judge it by what actually happened. The Division promised a summary about 30 days after phase one and a detailed report at about 90. Neither happened. The edibles were collected in October and November 2025; the full results reached the public in June 2026 (roughly eight months), after being described as imminent at four consecutive meetings, and after the program's reports were quietly consolidated into a single end-of-program document by a decision nobody at the forum could later locate. The transparency advocate here was heated enough to draw a rebuke, but his argument was never answered. It was outlasted.

About the demand itself, something nobody said in the room: what he asked the state to publish is something state law forbids the state to publish. Colorado's marijuana code (C.R.S. 44-10-204) makes individualized testing results (exactly the "complete COAs with batch and product codes" demanded here) confidential by statute. The Division couldn't grant that demand however it felt about transparency. What the same statute requires is the other half: testing records must be made publicly available in aggregated, de-identified form. And there's already a tool with a clock on it: the Colorado Open Records Act, under which any member of the public can request those aggregate records and the state has three working days to respond. So the productive version of this fight was never "everything, immediately" versus "trust us." It was: publish the aggregate data promptly, and if it doesn't appear, file a records request. In a year of these recordings, nobody appears to have taken that route. We're considering it ourselves.

Because note where things stand as we write, in August 2026: only some of the data has ever been released. The samples came off the shelves in October 2025, ten months ago. The full edibles report finally published around July, into the forum's sign-up-gated Drive folder; our own sweep of the open public record (2026-08-13) found no copy on MED's or CDPHE's open-web pages. And the aggregate data behind July's lab-comparison presentation, and the homogeneity spread data a manufacturer asked for in June, still haven't appeared. There are honest reasons fine-grained de-identification is hard with seven labs and nineteen manufacturers, and we've credited them. But at ten months and counting, with the statute's aggregate duty sitting right there, it is starting to look pretty bad.

What I'm genuinely curious about is why neither side reached for the law sitting between them. The regulators never mentioned the statute, and one sentence (the statute bars us from granting that) would have ended the argument. Why not say it? The obvious reading: naming the statute invites the follow-up, because the same section requires releasing the aggregate data, and a records request carries a three-day clock. Not naming it bought time. On the other side: why didn't a chief executive operating under this exact code reach for the records request himself? My best guess sits in the open questions below: aggregate data may simply have been useless for what he was actually after.

So the question that matters isn't who fumbled the statute. It's why the state appears to be stalling, and what it hoped to gain. Look at what the results said when they finally landed: only 47% of samples passed both label tests, a third of the failures were tangled up with testing exemptions, and lab-shopping got named from the podium. That's bad for the industry. It is just as bad for the state, because every bit of it happened inside Colorado's own regulatory system. Cannabis critics get to ask the regulators how things were allowed to get this sloppy, and an embarrassment for the industry is an embarrassment for Colorado, the state that sold the country on regulated cannabis. Seen that way, the delay stops being a mystery: an agency slow-walking its own bad report card is about the most human thing in government. And credit where due: they did publish it, self-indictment included, which is more than many agencies manage. But dread is not a legal basis. The statute's aggregate-disclosure duty has no exception for embarrassing data, and the Division denied on the record in April that investigations were holding the report while describing itself as "working with those companies" from the same dataset.

So before judging, we searched the entire 2026 public enforcement record ourselves: all 39 final administrative actions, all 20 health and safety advisories, all 7 industry bulletins, full text. Nothing in it traces to the surveillance program. Every 2026 action touching potency or testing exemptions grew from complaints opened before the shelves were ever sampled. Be careful what that timeline does and doesn't prove, though: cases that predate the sampling aren't necessarily unrelated to what the sampling found. They may be earlier windows onto the same behavior. One 2026 final action, from a complaint opened in 2024, charged a manufacturer with exactly the pattern the surveillance later flagged in a third of its failing samples: moving product on a testing exemption it hadn't earned. And the shelf failures themselves may be downstream of the lab-shopping the Division named in the same June presentation: a product that passed at an agreeable lab is exactly the product you'd expect to fail the state's retest. If enforcement and surveillance are seeing the same conduct through different windows, that supplies one more honest motive for the delay (releasing data that touches conduct already under investigation is genuinely delicate), and it makes the aggregate data more valuable, not less: it's the only way the public ever learns whether the patterns match. But the sweep surfaced a structural catch that matters more than the empty result: the program's own founding bulletin promised a failure ladder (quarantine, corrective action, recall, embargo), and its first two rungs produce no public record at all, by design. "Working with those companies" is the corrective-action rung, working as announced. From the outside, a state quietly doing exactly what it promised and a state doing nothing look identical. And final enforcement actions surface roughly eight months after they open, so a surveillance-driven case begun this spring wouldn't appear until 2027. The aggregate data is the one lever that would let the public tell the difference. That is why the release matters more than the blame.

The editor's open questions. There's a reading that makes the all-or-nothing position more coherent than it first sounds, and it leads somewhere worth being honest about. Are there operators in this industry that insiders widely suspect of gaming the testing system? The record gives real reasons to think so: the Division itself says sample adulteration "happens pretty consistently", a manufacturer's review put it in half of recent enforcement actions, and the regulator has named lab-shopping as an observed trend. Do the people in these rooms know who they suspect? My opinion, having sat in one of them: I think they do. That would explain exactly why aggregated, de-identified data held no interest for the advocate. Aggregate data can't name anyone. Is the withheld data protecting anyone in particular? Nobody outside the agencies can know. That is precisely the problem with dragging out the release. This is how things look; a statute that requires aggregate disclosure, aggregate data that stays unreleased month after month, and a room full of insiders who won't say names. Every month it drags on, it looks worse. The fastest way for the state to stop it looking that way is to release what the law already requires.

One thing the state got right in this meeting deserves saying just as plainly: asked to grade contamination failures on a curve, the answer was that the limits are health-based and every exceedance gets treated as what the limit says it is. And note that nobody in the room disputed it. Everyone, the meeting's fiercest critic included, conceded the point: "whatever the levels are set at, we have to adhere to them or else there's no bright line." A room that fought about everything else agreed on this. That's what a bright line is for.

Since this meeting

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

Resolved. The launch happened as designed: 45 edible samples collected across 16 stores in October and early November, per the January report-out. The test menu held, including this meeting's decision to drop potency from flower surveillance. And October's cancellation, made here on the record, is why the Division's archive lists no October 2025 recording.

Still in progress. The disclosure argument never ended: the promised confidentiality analysis never surfaced in a later recording we transcribed, the report-consolidation decision was contested in April, and advisory scope and CAPA approval continued into later drafting.

Put off or neglected. The 30/90-day reporting plan was not kept. The detailed results took about eight months. Who-pays-the-stores was left to licensees and stayed unsettled all year. And the sharp catch about sanitizing collection gear with isopropyl alcohol (itself an analyte on the residual-solvent panel) never returned in any later recording we transcribed.

What was brought up

The three launch documents [0:05:03]. The final collection SOPs (sanitized coolers, calibrated temperature loggers, sealed evidence bags, chain-of-custody manifests delivered to the reference lab); the sample-selection procedure (phased by product category, producers drawn by market-share strata, store-level substitution so an out-of-stock item doesn't break the process) [0:25:20]; and a draft industry bulletin telling licensees what to expect, including the inventory reason code stores would use when samples leave their shelves unpaid [0:30:21, 0:35:29].

The transparency demand [0:10:10]. An infused-products manufacturer pressed for full, near-real-time public posting of every result: "complete COAs with batch and product codes of everything that is tested, good or bad... within days, if not hours," on a website rather than bulletins. He argued the program is "a credibility building exercise" and that "there's just no such thing as partial transparency." The Division's stated plan: a public summary about 30 days after phase one ends and a detailed report at about 90 days, with health-and-safety advisories issued immediately on failures; unanalyzed confidentiality provisions may limit what can be shared [0:15:10]. The exchange grew heated enough to draw a chair's rebuke. The gap between "advisories now, reports later" and "everything, immediately" was not closed.

Who warns the people who already bought it [1:05:45]. The same stakeholder pushed the consumer-side timeline: the danger isn't future sales of a quarantined batch, it's people already holding contaminated product who keep using it while paperwork proceeds. The Division responded with a detailed explanation of what advisory date ranges mean: a batch's sale window is not when the state learned of the problem, and reading months-old batch dates as months of regulatory silence is, in the chair's words on the record, inaccurate [1:10:46, 1:15:52].

Graduated response, rejected [1:20:59]. A lab-sector participant suggested distinguishing barely-over-the-limit failures from order-of-magnitude ones, FDA-style, when choosing enforcement urgency. The state lab's answer: the limits are health-based, and every exceedance gets treated as what the limit says it is: a health risk, handled consistently. Notably, the transparency advocate agreed: "whatever the levels are set at, we have to adhere to them or else there's no bright line."

What was nailed down

• The test menu, by category [0:45:34]: flower and pre-rolls get microbials, pesticides, and elemental impurities (potency was dropped from flower surveillance); concentrates get potency and cannabinoids, residual solvents, mycotoxins, pesticides, and elemental impurities; edibles get potency, cannabinoids, homogeneity, microbials, and elemental impurities.
• Roughly 50 samples per phase, edibles first, launch targeted for the first week of October [0:50:38].
• Nobody pays the stores back: the MED collects samples without compensation; the state covers equipment, staff, and testing; the question of whether producers backfill stores was left to licensees [0:55:43].
• October's meeting was canceled on the record: the forum paused so staff could implement the launch, with a November meeting to be confirmed and a January restart for 2026 planning [1:26:04]. (This is why the Division's archive shows no October 2025 recording.)
• Interlab consistency was flagged as the coming year's priority topic: "we look forward to the data we're going to get from the surveillance study that can help inform that discussion" [1:31:09].

What was left open

• Public reporting of results: the 30/90-day plan versus real-time posting; the confidentiality analysis was explicitly not done yet [0:15:10].
• CAPA approval: a certification-body participant argued a corrective-action plan should require agency approval, not mere submission, with consequences for non-approval; the Division took it as a drafting change to consider [0:40:34, 1:00:43].
• Advisory scope: whether a failed batch's advisory should start store-specific (the Division's speed-first position, citing deli-style contamination scenarios) or batch-wide from the start (counsel for a manufacturer's position for pre-packaged goods) [1:00:43].
• Water activity in transport: raised as a sample-integrity variable the SOPs don't yet address, alongside a sharp catch: the collection SOP sanitizes with isopropyl alcohol, an analyte on the residual-solvent panel [0:20:16].
• Recall mechanics: the voluntary-recall/CAPA/embargo chain and who it actually protects, argued at length and continued into later meetings.

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.