Cannabis Breathalyzers: Why Roadside THC Tests Still Don't Exist

September 26, 2026

Photo by Kindel Media via Pexels.

Pull over a driver suspected of drunk driving anywhere in the United States and an officer can produce a number in under two minutes: blow into a handheld unit, get a blood alcohol concentration reading, and that number has been legally bulletproof since most states adopted per se BAC limits in the 1970s and 80s. Pull over a driver suspected of driving high, and that same officer is stuck with a toolkit that hasn't fundamentally changed in decades -- standardized field sobriety tests, a Drug Recognition Expert evaluation that takes 45 minutes to an hour, and if there's enough suspicion, a blood or urine draw that has to go to a lab and comes back days or weeks later. As of April 2026, zero states have a legally approved roadside THC breath test. Not one.

That gap isn't for lack of trying. Cannabix Technologies, Hound Labs, a Virginia Commonwealth University forensic science team, and the National Institute of Standards and Technology have all spent years and, in some cases, real federal grant money chasing a device that does for THC what Robert Borkenstein's Breathalyzer did for alcohol back in 1954. The holdup isn't engineering laziness or a shortage of capital. It's that THC in the human body behaves nothing like ethanol, and building a sensor that can detect a molecule is a much easier problem than building one whose reading means something specific and defensible about impairment at the moment of the stop. This is very much a mid-term story. The chemistry is getting genuinely good. The legal and regulatory scaffolding it needs to sit inside is still years from being built.

Why Breath Testing THC Isn't Just a Smaller Version of the Alcohol Test

Photo by Łukasz Promiler via Pexels.

The alcohol breathalyzer works because of a tidy piece of physical chemistry called Henry's Law: at body temperature, alcohol dissolved in blood equilibrates with alcohol vapor in the air of the lungs at a stable, predictable ratio. Measure the vapor, apply a partition coefficient, and you get a defensible estimate of blood alcohol concentration. That relationship has been validated so many times, in so many courtrooms, that it's essentially settled science. It's the reason a $500 handheld device can produce evidence that survives cross-examination.

THC doesn't play by those rules. It isn't dissolved in blood in a way that equilibrates cleanly with lung air. What shows up in exhaled breath after cannabis use is mostly residual particulate matter -- tiny droplets of oil, resin, and combustion or vaporization byproducts that coat the mouth, throat, and airway lining after smoking or vaping. That's a fundamentally different transport mechanism than a gas exchanging across the alveolar membrane, and it means there's no equivalent of Henry's Law waiting to be discovered here. A 2023 federally funded report from NIST and the University of Colorado Boulder reviewed the existing science and concluded, bluntly, that the evidence does not support the idea that a single breath THC measurement can reliably indicate recent use, let alone impairment.

Timing compounds the problem. THC breath concentrations spike within minutes of smoking or vaping and then fall off a cliff -- researchers have documented drops of one or two orders of magnitude within just a couple of hours. Blood alcohol also declines over time, but predictably, at a roughly linear rate that toxicologists can back-calculate with confidence. THC's decay curve is steep, fast, and variable between individuals depending on how they consumed it, how much lung tissue retained residue, and basic physiology. That means the exact minute a breath sample gets collected matters enormously to what the number means, in a way it simply doesn't for alcohol.

So the question facing every company in this space isn't just can you detect THC in a breath sample. Mass spectrometry can do that at trace levels already. The question courts and legislatures actually care about is whether a positive reading tells you anything meaningful about whether the driver was impaired in that moment, thirty seconds ago, versus having smoked at a party the night before. Right now, the honest scientific answer is: not reliably, not yet.

Cannabix Technologies: Workplace Traction, Roadside Still Distant

Photo by cottonbro studio via Pexels.

Cannabix Technologies (CSE: BLO) has been the most publicly aggressive player chasing this problem, and its Marijuana Breath Test, or MBT, was originally pitched as a dual-use tool for both roadside enforcement and workplace screening. In early March 2026, an independent, peer-reviewed study published in the Journal of Analytical Toxicology gave the company a genuine scientific win: it confirmed the MBT can detect delta-9 THC within roughly a four-hour window of use, at concentrations above 5 picograms per liter of breath. That's a real, third-party-verified result, and it matters -- it's the kind of data point regulators and defense attorneys will eventually want to see repeated and scrutinized.

To turn sensitivity into something usable in a forensic or legal chain of custody, Cannabix partnered with Omega Laboratories, a forensic drug testing lab with more than 25 years in the field, pursuing what's called a Laboratory-Developed Test Method pathway -- essentially, a route to validating the device's output as something a lab, and eventually a court, can stand behind.

Here's where the story gets more complicated than the press releases suggest. Cannabix's actual commercial deployment news in 2026 -- a September 23rd announcement of a device going into use at a manufacturing plant in Alabama -- was for BreathLogix, the company's alcohol screening tool, not the THC device. Likewise, when Cannabix secured a spot on NHTSA's conforming products list in 2026, that approval applied to the alcohol product. The marijuana breath test, the one everyone actually wants, still doesn't have that kind of regulatory sign-off anywhere.

Put those threads together and a reasonable read emerges: Cannabix is a real company doing real breath-detection work, and it's proving out its underlying sensor technology commercially -- just through alcohol and workplace products, where the legal bar is lower and the science is already settled. The roadside THC product remains, as of today, a peer-reviewed detection method in search of a regulatory pathway, not a device any police department can legally deploy.

Hound Labs: The Most Credible Science, the Quietest Company

Photo by Ivan S via Pexels.

If Cannabix is the loud one, Hound Labs is the quiet one -- and on paper, its science is arguably the more impressive of the two. Hound has published data claiming picomolar-level sensitivity, detecting THC in breath down to parts per trillion, with a reported cutoff around 20 picograms per milliliter. That's a finer resolution than Cannabix's published 5 pg/L threshold, at least on paper, and it's the kind of number that suggests genuinely sophisticated sensor engineering.

Both companies have gotten a cannabis breathalyzer into commercial use for workplace settings -- employers doing pre-shift or post-incident screening, essentially, rather than law enforcement doing roadside stops. Neither has a device cleared for law enforcement use anywhere in the United States. That's worth sitting with: after years of development and real investment from both companies, the furthest either has gotten commercially is the workplace, not the road.

What's odd about Hound specifically is how thin its public trail has gotten. Industry tracking sites and cannabis tech watchers note that Hound's public updates have been sparse since roughly 2020, and its current operational status as of 2026 is genuinely unclear from the outside -- no clean confirmation of active fundraising, product shipments, or new peer-reviewed data in the way Cannabix has managed to generate. That's not proof the company has stalled. Companies go quiet for all kinds of reasons, including simply not needing to feed a press cycle while they work. But it is a real data point, and it cuts against Hound in any comparison with a competitor that's been visibly, if selectively, announcing wins.

The broader lesson here applies to the whole sector, not just Hound: a laboratory demonstrating picomolar sensitivity under controlled conditions is not the same problem as a handheld device that survives being used at 2 a.m. on the shoulder of a highway, in cold or humid weather, by an officer who isn't a chemist, and that then has to survive a defense attorney's cross-examination months later. Technical credibility and commercial follow-through are two different races, and right now it's genuinely unclear which company, if either, is actually winning the second one.

The VCU Inhaler and NIST's Federal Push

Photo by Jorge Chan via Pexels.

The most scientifically interesting recent development didn't come from either commercial player -- it came from an academic lab. A study published April 8, 2026, led by then-Virginia Commonwealth University forensic scientist Emanuele Alves and partly funded by the Justice Department, described a device that looks and works nothing like the Cannabix or Hound sensors. It resembles an inhaler, built around 3-D printed cartridges using a Fast Blue dye paired with a gelatin colorimetric system -- meaning the detection method is a visible color change rather than an electronic sensor reading. That's a fundamentally cheaper, simpler approach, at least in concept, than the mass-spectrometry-adjacent sensing Cannabix and Hound are pursuing.

The colorimetric experiments produced what the research team called foundational data -- evidence the device can detect delta-9 THC, along with CBD and CBN, across multiple sample matrices. That's an early-stage result, not a finished product. Alves has since left VCU and could not be reached for comment on the project's future, though the published work states a patent has been filed for a next-generation prototype. That's worth reading for exactly what it is: a signal of early intellectual property protection, not evidence of an imminent commercial launch. VCU toxicologist Michelle Peace remains at the institution and represents the kind of ongoing forensic expertise that could carry this line of research forward even with its original lead researcher gone.

Running in parallel, and arguably more important for closing a real gap in the field, is work out of NIST and CU Boulder's Anschutz campus. A July 2025 study from that team achieved something no one else has published: the first documented detection of THC in breath following edible ingestion, paired with driving-simulator assessments of impairment. That matters because every device discussed so far -- Cannabix, Hound, the VCU inhaler -- is fundamentally built around detecting residual particulate from smoking or vaping. Edibles don't leave that same oral and lung residue, so a smoke-optimized breath sensor could plausibly miss a driver who's genuinely impaired but consumed cannabis as a gummy rather than a joint. As edibles keep gaining share of the legal cannabis market, that blind spot gets more consequential, not less.

Tara Lovestead, one of the NIST researchers on that work, has been explicit about what still stands between any of this and a courtroom: even a technically working sensor is going to need formal accuracy standards -- the kind of standardized, replicable performance benchmarks NHTSA eventually built around alcohol devices -- before any THC breathalyzer could be trusted as legal evidence.

What Has to Happen Before Any of This Reaches the Roadside

Photo by sergeitokmakov via Pixabay.

History offers a useful, if sobering, timeline here. Rolla Harger built the Drunkometer in 1938. It took until 1954 for Robert Borkenstein's Breathalyzer to standardize the technology, and it took until the 1970s and 80s for states to widely adopt per se blood alcohol limits and for NHTSA to formally standardize approved devices through its conforming products list. That's roughly forty years from working prototype to a legally uniform national enforcement standard -- and alcohol had the advantage of Henry's Law giving it a clean, provable scientific foundation almost from the start.

THC doesn't have that advantage yet, and it's reasonable to expect a comparably staged, multi-step sequence before a device reaches the roadside anywhere: first, a validated correlation between a specific breath biomarker and actual, measurable impairment (not just presence). Second, a body like NHTSA setting formal accuracy standards a device has to meet, the way it eventually did for alcohol breathalyzers. Third, state legislatures writing statutes -- either per se numeric limits or presence-based rules -- that actually reference a breath standard rather than blood or urine. Fourth, that evidence surviving real court challenges from defense attorneys who will, correctly, point to the unresolved science around what a THC reading actually proves about impairment.

Complicating all of this is that there's no single finish line to cross, because state DUI law around cannabis is already a patchwork. Some states use per se THC blood limits, commonly around 5 nanograms per milliliter. Others use zero-tolerance standards where any detectable THC is a violation regardless of impairment. Others rely entirely on a Drug Recognition Expert's trained observational judgment, with no numeric threshold at all. A breathalyzer that worked perfectly tomorrow wouldn't slot into one clean national framework -- it would have to be individually adopted, validated, and litigated across dozens of different legal regimes.

There's also a harder scientific counter-case that alcohol never had to contend with: even among researchers who study cannabis impairment closely, there's genuine, ongoing disagreement about how tightly any single biomarker -- breath, blood, or otherwise -- tracks with actual driving impairment. That's not a solvable engineering problem the way sensor sensitivity is. It means a device could someday detect THC in breath with excellent precision and still fail to hold up as impairment evidence in court, because detection and impairment are scientifically different questions.

All of which points to a real business opportunity that simply isn't open yet. Law enforcement agencies nationwide are actively looking to replace slow, expensive blood-draw workflows, and whichever company gets there first with a legally admissible device stands to capture that entire market. But that market opens on regulatory time, not product-launch time -- and regulatory time, on this particular question, has historically run in decades.

The workplace is where this technology is actually getting tested right now, even if nobody's framing it that way. Every deployment at a manufacturing plant, every pre-shift screening program, every batch of field data Cannabix or Hound quietly accumulates from employers is real-world validation happening well before anyone asks a court to trust these devices with a DUI conviction. That's arguably the smart path -- workplace screening has a lower evidentiary bar than criminal law, and it gives these companies years of field performance data to point to when they eventually do approach NHTSA and state legislatures.

Don't expect a single winner to emerge and declare victory. The more likely mid-term outcome is a slow, unglamorous convergence: NHTSA or an equivalent body eventually publishing accuracy standards, state legislatures rewriting DUI statutes to reference breath rather than blood, and courts gradually building precedent case by case, the same lagging pattern alcohol testing went through between Harger's 1938 Drunkometer and the standardized enforcement regime of the late 1970s. If that history rhymes even loosely, this is a most-of-the-decade story, not a next-product-cycle story.

One thread is worth watching more closely than the corporate press releases: the NIST and CU Boulder edibles research. A breathalyzer that only catches smoked or vaped THC residue is solving yesterday's cannabis market. Edibles keep taking share from flower, and a device with a permanent blind spot for ingested THC isn't a complete solution no matter how many picograms per liter it can detect. Whoever eventually closes that gap -- not just the smoke-and-vapor problem, but the edibles problem too -- is the one actually building the tool law enforcement needs, rather than the tool that's easiest to build first.