Every good legal thriller needs a twist, and this one’s a doozy: The DEA — the agency that’s spent over 50 years playing prohibition’s final boss — is now the one arguing marijuana doesn’t belong in Schedule I anymore, while the seven groups it hand-picked to participate in the hearing are the ones fighting to keep it there. It’s the regulatory equivalent of watching the Empire Strikes Back reveal, except instead of “I am your father” it’s “actually, we agree with HHS now.”

After 11 days of hearings, a judge who couldn’t stop asking witnesses to define “marijuana” like it was a trick question, and briefs that spent as much time trying to disqualify each other’s experts as arguing the actual science, the closing arguments are in. Here’s what they actually say — and why the fight over which medical-use test applies might end up mattering more than anything said under oath.

Background

The government defends a rule proposed under the Controlled Substances Act to reschedule marijuana from Schedule I to Schedule III. Not just medical marijuana (which, broadly speaking, was moved to Schedule III earlier this year) but also adult use, often called “recreational” marijuana.

Seven interested parties — Smart Approaches to Marijuana (SAM), the states of Nebraska, Idaho, and Indiana, the Tennessee Bureau of Investigation (TBI), the National Drug and Alcohol Screening Association (NDASA), DUID Victim Voices, and Dr. Kenneth Finn — oppose it (Louisiana withdrew before the hearing started). Briefs were filed around August 18, and the full corrected transcript (2,533 pages) was released around August 25. DEA ALJ Derek Julius has now closed the record and will issue a recommendation to DEA Administrator Terry Cole.

That recommendation is advisory only: Cole can adopt it, reject it, or move marijuana to a schedule other than the one either side proposed, and his ultimate decision — not the ALJ’s recommendation — is what would be subject to judicial review.

Here’s who testified for the parties and witnesses that recur below:

• Government: Dr. Dominic Chiapperino (FDA, director of Controlled Substance Staff)
• NDASA: Patrice Kelly (34-year DOT veteran) and Mary Jo McGuire (NDASA executive director)
• SAM: Dr. Bertha Madras (Harvard psychobiologist, former ONDCP official) and Dr. Luli Akinfiresoye (a sitting DEA pharmacologist)
• Opposing States (Nebraska, Idaho, Indiana): Dr. Deepak D’Souza (Yale psychiatry)
• Dr. Kenneth Finn: Dr. Finn himself
• Dr. Phillip Drum (appearing pro se): Dr. Drum himself and Dr. Karen Randall (Colorado emergency physician)

With that scope and cast in mind, five fights dominate the record:

1. Is the new “two-part test” for medical use even a valid test — and is that question off the table?

The core legal mechanism for rescheduling under the federal Controlled Substances Act is HHS’s determination that marijuana has a “currently accepted medical use” (CAMU), a prerequisite for leaving Schedule I. Historically, that finding required a “five-part test” from a 1988 DEA notice (following Alliance for Cannabis Therapeutics v. DEA):

• Known/reproducible chemistry;
• Adequate safety studies;
• Adequate efficacy studies;
• Acceptance by a body of qualified experts; and
• Widely available scientific evidence.

In 2023, HHS switched to a new “two-part test” — used for the first time on marijuana — that asks only whether there’s (1) widespread practitioner-supervised use under state programs and (2) “credible scientific support” for at least one condition.

The government’s position: A 2024 Department of Justice Office of Legal Counsel (OLC) opinion already blessed the test — finding the old five-part test “impermissibly narrow” — and, according to DEA, that opinion binds the department, including the ALJ in this case. In DEA’s view, the hearing exists to apply the test, not relitigate it.

The opposition’s position: The two-part test is dramatically less rigorous than the five-part test it replaced. It drops the requirements for reproducible chemistry and for acceptance by a body of experts, and Dr. Chiapperino himself conceded under cross-examination that marijuana likely would not have passed the old five-part test. He also confirmed that his team started its review under the old regime in late 2022 and was only formally told to switch tests in July 2023, months later (notably, perhaps, under a new administration). NDASA/Finn’s counsel pushed further, arguing that “credible scientific support” is so loosely defined — no requirement of study size, replication, recency, or even a well-designed trial — that virtually any Schedule I drug could pass it.

2. Does the “currently accepted medical use” finding actually match how marijuana is used medically?

If we accept for the purposes of this article the two-part test, there was substantial testimony on both sides about how marijuana is actually used in practice. Of roughly 15 conditions that met HHS’s Part 1 threshold (widespread state-program use), FDA only ran Part 2 scientific-support analysis on seven conditions — and found sufficient support for just three:

• Pain (particularly neuropathic pain);
• Anorexia associated with a medical condition; and
• Chemotherapy-related nausea/vomiting.

PTSD, anxiety, epilepsy, and inflammatory bowel disease were tested and did not pass.

The government’s position: The Controlled Substances Act requires finding just one qualifying medical use to justify moving a drug out of Schedule I, and marijuana cleared that bar for three conditions using a scientifically defensible, multi-source methodology (national surveys, a contracted University of Florida literature review, FDA’s own review, and state program data from Maryland and Minnesota). DEA’s closing brief adds a numeric argument on the closely related “accepted safety under medical supervision” prong: Over 30,000 practitioners now treat more than six million patients across 43 jurisdictions, which DEA argues demonstrates safety through sheer scale of supervised use.

The opposition’s position: Dozens, “if not more than 100,” conditions are treated with medical marijuana in state programs that never passed Part 1, Part 2, or FDA approval — meaning the CAMU finding doesn’t reflect the reality of who’s actually using marijuana and why. Opposing states’ counsel also drew out that of the states with medical marijuana programs, only 17 of 38 require any medical-board review to approve a condition, only 10 require practitioner training, and only 20 track adverse events at all — undercutting the claim that “widespread practitioner-supervised use” reflects genuine medical oversight rather than loosely regulated access. Dr. Finn (a Colorado/Arizona pain physician who registered as a medical marijuana patient himself as a test case) testified he was approved for “severe pain” in under 60 seconds via telehealth, with no review of his actual pain level, medical records, prior treatments, or drug interactions. And in Colorado, three doctors made roughly a quarter of all medical marijuana recommendations statewide in 2022, a concentration pattern he compared to opioid-era “pill mills.” DEA’s own closing brief tried to turn some opposition witnesses’ testimony back on them here: It points out that SAM’s own addiction expert, Dr. Bertha Madras, conceded there’s scientific evidence of efficacy for neuropathic pain (while calling it weak), and that Dr. Finn himself acknowledged he’d previously recommended marijuana to his own pain patients and believes there’s likely a therapeutic benefit — with his real objection being dosing and quality control rather than medical value itself.

3. How should the tribunal weigh marijuana’s relative abuse potential and harms against comparator drugs?

FDA’s analysis compared marijuana’s abuse potential and harm rates against heroin, fentanyl, oxycodone, hydrocodone, cocaine, ketamine, benzodiazepines, and — notably — alcohol, an uncontrolled substance included because of marijuana’s similarly widespread recreational legal access.

The government’s position: By nearly every harm metric — overdose deaths (marijuana lowest among all comparators, 2012–2021), risk of severe withdrawal (milder than alcohol or benzodiazepines, more like tobacco), and psychological dependence severity (skewing mild versus moderate/severe for the comparators) — marijuana poses less absolute risk than the drugs it would join in Schedule III, let alone the Schedule I and II drugs it’s leaving behind. DEA’s closing brief adds that “the vast majority of individuals who use marijuana do so in a manner that does not result in dangers to themselves or to their communities.”

The opposition’s position: The comparison is incomplete and sometimes cuts the other way. On emergency-department visits, marijuana actually had the second-highest utilization-adjusted rate among the substances studied — higher than alcohol, though lower than cocaine (no opioid comparator was available in that particular dataset). NDASA/Dr. Finn’s counsel also pushed on categories the formal eight-factor write-up doesn’t address at all: associations with psychosis/schizophrenia, depression, cardiac events, and reduced cognitive function from chronic use. Dr. Chiapperino acknowledged these weren’t documented in the analysis, though he said some were informally considered elsewhere. Dr.  D’Souza (representing the opposing states) testified that in Scandinavian registry studies, 50-75% of people diagnosed with cannabis-induced psychosis were later re-diagnosed with schizophrenia; he also conceded under the government’s cross that most people who use marijuana do not experience psychosis or other negative mental-health consequences. DUID Victim Voices and Dr. Finn’s closing brief presses this hardest, arguing DEA “systematically ignored or underweighted” psychosis, schizophrenia, suicidality, cardiovascular harm, and impaired driving, and separately attacks the alcohol comparator as pharmacologically mismatched and not a standard CSA scheduling comparator in the first place.

4. Would rescheduling gut federal workplace drug testing?

NDASA’s witnesses — Kelly, a 34-year DOT veteran who co-authored the agency’s governing testing regulation (49 CFR Part 40), and McGuire, NDASA’s executive director — testified that Reagan’s 1986 executive order authorizing mandatory federal drug testing only covers Schedule I and II substances. Move marijuana to Schedule III, they argued, and it falls outside that authority entirely, potentially ending mandatory testing for pilots, air traffic controllers, truck drivers, train engineers, transit and pipeline operators, and other safety-sensitive federal roles. Kelly put the number of affected workers at roughly 10 million, covered by more than seven million tests a year, the largest regulated testing program in the world. A more granular, practical problem sits underneath that: The physician gatekeepers who verify positive drug tests against valid prescriptions (medical review officers) normally call a pharmacist to confirm a prescription is real; under Schedule III they’d have to call a marijuana dispensary instead, and dispensaries aren’t required to disclose that information the way licensed pharmacies are, nor do most medical-marijuana recommendation cards even list a prescribing physician. McGuire also testified to a personal case: Her own son developed psychotic episodes after starting recreational marijuana use as a teenager, held a stable job for years under a mandatory-testing requirement, then relapsed into heavy high-potency use once he moved to a job without testing. He is now serving a 10-year sentence for armed robbery.

The opposition’s position: The NPRM never addressed this consequence at all. NDASA’s closing brief narrows this to a concrete stakes argument: If marijuana moves to Schedule III, federal authority to mandate testing for it could lapse for air traffic controllers, Secret Service, pilots, truck drivers, transit operators, and other DOT-regulated positions.

Where it stands: This wasn’t substantively rebutted by the government in the portions of the transcript covering NDASA’s case; the objections raised were mostly about the scope of the witnesses’ qualifications and the relevance of testimony about a different, already-issued attorney general order, not about the underlying legal-authority problem itself.

5. Are state marijuana programs actually protecting patients?

This theme ran through nearly every opposition witness, but two were especially granular. Dr. Drum compared FDA’s own dronabinol (oral THC) label — which flags 28mg/day as producing significant CNS symptoms — against California’s own regulations permitting up to 2,000mg of THC per package for medical concentrates. He cited investigative reporting and state recall databases showing marijuana-product recalls (for mold, pesticides, and mislabeled potency) are almost always “voluntary” and can take months to issue even for products found to contain a potentially fatal mold species. Dr.  Randall, a Colorado emergency physician appearing for Dr. Drum, testified that the Prescription Drug Monitoring Program (PDMP) — the system nearly all U.S. prescribers rely on to check whether a patient is already on a dangerous drug combination — contains no marijuana data at all, so a treating physician has no way to know if a patient is using it, at what dose, or from where, and described “looping” (visiting multiple dispensaries to exceed per-visit purchase limits) as a known, poorly tracked problem in Colorado.

The opposition’s position: State programs are regulatory in name only. Dr. Randall toured a dispensary and testified that its “medical” and “recreational” counters sold identical products, differentiated only by counter color; Dr. Drum testified that mislabeled potency (edibles off by up to 20-30% from stated dosage in some studies) and packaging that closely mimics children’s candy brands are widespread and effectively unpoliced.

The government’s position: This point largely went unrebutted on the specifics in the portions I reviewed; cross-examination of these witnesses focused more on scope-of-expertise objections than on disputing the underlying regulatory-failure evidence.

Where the Closing Briefs Left It

DEA’s closing position sharpened into three planks: (1) Marijuana has a CAMU under the two-part test; (2) it has “accepted safety for use under medical supervision” — leaning on the 30,000-practitioner, six-million-patient figure described in Issue 2; and (3) its abuse/dependency profile fits Schedule III better than Schedule I or II. DEA’s brief reportedly argues the two-part test needs no independent defense because opposing witnesses undercut their own case — pointing to Dr. Madras’s and Dr. Finn’s partial concessions described above. DEA also reportedly attacked the reliability of some opposition witnesses, alleging at least one violated a sequestration order by reading other witnesses’ transcripts before testifying (this appears to reference Dr. Finn, who candidly disclosed on the record that he had read days 1 and 2 of the transcript because they were emailed directly to him as an interested-party representative; the ALJ declined to exclude his testimony on that basis, ruling it would go to weight rather than admissibility).

Dr. Drum’s closing brief reportedly takes the most categorical position of any opposition party: that marijuana, given the range of delivery formats now marketed (topicals, suppositories, inhalers, edibles, and more) with unknown per-format dosing, isn’t a “medicine” in any coherent sense, and that state programs aren’t actually protecting the public despite claiming to regulate it — consistent with his testimony described above.

Bottom Line

The case is less a fight over whether marijuana is “dangerous” in the abstract than over methodology: which test applies, whether that test’s legal footing is even settled, how narrowly or broadly “medical use” should be read against real-world state programs, which harms count and which don’t, whether the government’s own witnesses actually support its case, and how stale or incomplete a record can be before it undermines the rule built on it. None of these questions were resolved during the hearing itself — they’re now sitting with the DEA ALJ, who will issue a recommendation to the DEA administrator. Cole makes the final call and isn’t bound by the recommendation; under 21 CFR 1316.66, designated parties get 20 days after the recommendation to file exceptions with citations to the transcript and exhibits before the record goes to the administrator for final action.

Conclusion

Strip away the five fights above and one pattern holds across all of them: This hearing was never really a dispute about whether marijuana is dangerous in some abstract sense. It was a dispute about whose measuring stick gets used — which test for “medical use,” which comparator drugs, which data cutoff date, which witnesses count as credible, and which harms get counted at all. The government built its case on a new methodology that produces a different answer than the old one did, without much dispute that the underlying pharmacology hasn’t changed since the last three times DEA looked at this question. The opposition spent 11 days trying to show that methodology is result-driven or incomplete — and in at least one instance, got a sitting DEA employee to say things that cut against DEA’s own position, unchallenged by the government’s own cross-examination.

None of that was resolved in the hearing room. It didn’t have to be. An administrative law judge’s job here isn’t to settle the marijuana debate; it’s to build a record and hand up a recommendation. That recommendation now goes to Cole, who isn’t bound by it and can land anywhere from keeping marijuana in Schedule I to moving it to Schedule II or III. Whatever he decides becomes the actual final agency action — the thing courts can review, and almost certainly will, given that lawsuits over this rescheduling effort are already underway before it’s even finished. The hearing settled the record. It didn’t settle the argument.

We’ll stay on top of developments in marijuana rescheduling so you don’t have to.

Thanks for stopping by.

Inside the DEA Marijuana Rescheduling Hearing: The Five Fights That Will Decide the Case