The August 7 testing meeting: what got settled, what got raised, what's still open

Colorado's Science & Policy Forum (the working group where the Marijuana Enforcement Division and the state health department hash out testing science with labs and industry) met on Friday, August 7, 2026. The state published the video afterward. We transcribed the full 77 minutes and read it, and this is what happened, sorted three ways: what got settled, what got raised, and what nobody could answer.

Everything below comes from that recording, with timestamps so you can check any of it yourself. This article is part of our testing overhaul story. Since it was published, we've covered the entire year of forum meetings that led here, meeting by meeting, in our Science & Policy Forum archive, with a year-in-one-read analysis of the questions that kept coming back. This meeting is where several of those year-long threads ended up.

The headline: third-party sampling is off this forum's table

The single biggest proposed change in Colorado marijuana testing has been the idea that businesses should stop collecting their own compliance samples: an independent party should pull them instead. It is the fix most often floated for sample manipulation, and it was the first of three subtopics this forum took up under its "interlab consistency" umbrella this year.

It will not come from this forum. At 6:04 in the recording, the forum was told that it had evaluated "whether or not the science and policy forum would make a recommendation for implementing third party test batch collection as a requirement in MED rules," and that "following the June meeting… there was consensus not to move forward with that as a science and policy recommendation." (That June decision is covered in the archive, along with the full arc of the proposal from January's scheduling through March's cost fight to the room declining it.)

Read that precisely, because the distinction matters: the forum decided not to recommend it. The forum advises; it does not write rules. Some related sampling suggestions, the meeting was told, "are being considered as part of the rulemaking process, independently of the science and policy forum." So third-party collection is not dead as an idea in Colorado. It is dead as a recommendation from the body that spent this year studying it. Whether any version of it appears in proposed rule text is a separate question, and an open one.

What else got settled

The state reviewed its own method-validation rules and found them sound. The health department's laboratory certification team ran a redline review of the method validation criteria in rule. Their conclusion, at 17:22: "the current criteria is comprehensive and sound," though "additional clarification and guidance would be helpful." No overhaul of validation requirements is coming out of this review, only sharpening.

Sample manifests now show where the sample came from. A change completed in the month before the meeting adds the source package's size to the manifest, the paperwork that travels with every test sample. From 68:01: if a 100-gram package yields a 2-gram test batch, the lab now sees both numbers. It applies only to test-batch transfers to testing facilities, not to ordinary business-to-business transfers. Asked whether the system therefore verifies that a big enough sample was pulled, the answer was no: it documents what was reported, and nothing rejects an undersized sample.

Could lab wiggle-room stack up past the label's legal limit? A clear answer. A participant asked whether a lab's 20 percent quality-control tolerance could let a reported potency result land outside the 15 percent variance the rules allow (62:58). The reference lab's answer: no. Anything beyond about 10 percent variance in those control checks triggers "a full investigation," repeat extractions and duplicate injections. Note the scope: that describes the state reference lab's own practice, not a rule binding every private lab.

Surveillance testing moves to rulemaking. The program that buys products off the shelf and retests them produced recommendations, and those have become proposals inside the Division's ongoing rulemaking. The forum is deliberately pausing its own surveillance discussions rather than running a parallel conversation (7:32–8:20).

What got raised

The state's own scientists walked through what their reference lab does that the rules do not require, and invited industry to say whether writing any of it into regulation would help:

• A standard way to establish detection limits. Labs currently pick their own approach for verifying limit of detection and limit of quantitation (the smallest amount a method can reliably see, and the smallest it can reliably measure); a shared reference method was floated as a consistency fix (18:23).
• Calibration rules beyond the single number in the rule. The rule requires an R² of 0.995, a single statistical score for how cleanly a lab's calibration line fits. The state lab holds itself to more: retention times stable within 0.05 minutes, fresh calibration curves every two weeks or after any maintenance, dated standards, quality-control checks bracketing every ten samples.
• Trend tracking, which the rule never mentions. The state lab watches its own results drift over months. "Trend tracking is something that we certainly encourage, but it's definitely not addressed by the rule" (34:22).
• Matrix-specific controls. The rules do not distinguish between controls made in solution and controls made in an actual product matrix (34:50). A lab director pushed back usefully on how far that could go: flower is easy, but concentrates and edibles fracture into endless subtypes: "a brownie is a brownie, but a brownie that has walnuts and chocolate chips is different," and specifying each could mean running twenty control samples per batch.
• Everything that happens before the sample reaches the instrument. A manufacturer raised sample handling in transport and prep (including, bluntly, food in laboratories). The response from the state, at 46:35: "on the record, no one should ever have food in the laboratory," a point its inspectors assess on site. Sample-handling protocols are the September agenda.
• A live compliance limbo for edibles. A participant reported that a standard operating procedure covering non-infused and partially infused edibles changed before the corresponding rulemaking, leaving labs unsure how to test those products (62:13). The state took the issue offline to review it. If you make partially infused products, this one is worth watching.

What was left open

The big one: labs agree with each other, then disagree with the state. A lab director laid out the puzzle directly (52:43). In proficiency tests and round-robin studies (the same prepared sample sent to every lab), Colorado labs land close together. But surveillance testing (where the state buys products and retests them) has shown large gaps between what a lab originally reported and what the state reference lab found (the July meeting had put a number on it: lab-passed samples passed the state's retest barely more than half the time). "What are we doing that you guys are doing, aren't doing, or what are you doing that we're not doing?" he asked. "It makes us nervous."

The answer, at 53:38, was: "I don't think we have an answer to that."

The state offered two partial explanations without claiming either settles it: the surveillance sample is not the same physical sample the original result came from (54:45), and storage conditions (moisture loss in edibles over time, for instance) might account for some of it. Nobody in the room claimed the gap is understood. For anyone who reads potency numbers on a label, this is the most consequential unresolved question in Colorado cannabis testing, and it is unresolved on the record.

Whether any of this actually fixes consistency. A lab representative was frankly skeptical that tighter definitions of detection limits would move the needle on the inconsistencies that matter, particularly pesticide results. Without "that golden standard of the method that every lab should use," he said at 59:51, "we are going to have differences", and labs cannot see each other's methods to find out where the divergence starts.

Rule or guidance? Repeatedly asked, never settled: should these tightened expectations become regulation, or stay in the audit checklist as guidance? The state named the tension itself: write rules too specifically and they break for some test or some matrix; leave them general and labs diverge.

Manifest edge case. If a test batch is created and then adjusted, does the manifest show the adjustment or only the original number? Flagged for follow-up, unanswered.

What happens next

The forum meets again September 4, 2026 (held firm despite falling on the Friday before Labor Day weekend), and the topic is sample collection procedures, which is where several of the raised issues are supposed to land. We're collecting reader questions for that meeting. If there's something you'd want asked, send it. A separate technical forum on the Aspergillus method is being scheduled, with virtual and in-person options and a registration link to come; the metals method that was also slated has been pushed to a later date. The forum is collecting topic proposals now for next year's cycle.

We will be watching all of it. If you were in that meeting and we have mischaracterized something you said, or you want to correct a name, write to editor@colorado-medical-cannabis.org. Corrections from people who were in the room are the best kind.

Updates since this article published

• The Aspergillus technical forum got its date: August 20, 2026, hybrid, announced to registered forum participants (this site's editor is one) on August 11.
• The "labs cannot see each other's methods" complaint moved, by half. CDPHE has now publicly posted five of its reference-lab methods: potency, heavy metals, salmonella/STEC, total yeast and mold, and Aspergillus, redacted only for vendor details. The state's side of the method curtain is open; the private labs' methods remain unpublished, and the still-in-validation pesticide and solvent methods are yet to appear.
• The year behind this meeting is now fully covered: nine meeting-by-meeting summaries and a cross-year analysis live in the forum archive.