Cannabis DUI Laws: Why THC Blood Limits Don’t Measure Impairment

Blood THC limits can’t tell an impaired driver from a sober daily user, and roadside tests flag people who aren’t high. Nearly a decade after VICE’s “Stoned Driving” episode, the science is clearer. The law, especially for medical patients, mostly isn’t.

A driver on Xanax and a driver with a medical cannabis card can be stopped on the same road, on the same afternoon, and face two different laws: one has to be shown to be impaired, the other only has to test positive.

That isn’t true everywhere. But according to the Governors Highway Safety Association, 18 states punish THC by presence or by number, whatever the driver’s actual condition. Benzodiazepines, which carry a larger and more consistent crash risk, are generally prosecuted on evidence of impairment.

The gap matters because the number those laws rely on doesn’t measure what the law says it does.

How High Is Too High to Drive?

Nearly ten years ago, VICE’s classic Weediquette aired an episode called “Stoned Driving.” Its central case was a medical cannabis patient in Washington who had switched from prescription opioids to a joint a day for chronic pain, was pulled over for speeding, and tested at 22 nanograms of THC per milliliter of blood, more than four times the state’s legal limit.

The episode asked how stoned is too stoned to drive, and whether the law has any way to tell. At one point, host Krishna Andavolu hit a bong and drove a closed test course. He drove slowly, knocked over cones on the slalom and reacted late when a ball rolled into his path.

A decade of controlled trials, crash studies and court rulings later, part of that question (“how high is too high?”) has a fairly clear answer. The rest of it, especially what it means for patients, is roughly where it was ten years ago.

There’s Still No Clear Standard for Blood Testing in Cannabis DUI

In Washington, under RCW 46.61.502, a driver with 5.00 ng/mL or more of THC in blood within two hours of driving is guilty of DUI regardless of how they drove, and being legally entitled to use the drug, as a medical patient is, is no defense. Other states use 2 ng/mL, and some treat any detectable amount as an offense.

The case against a numeric line was already on record when VICE was filming. In 2016 the AAA Foundation for Traffic Safety, hardly a cannabis lobby, analyzed data from drivers arrested for DUI and tested thresholds from 1 to 10 ng/mL. None produced acceptable agreement with roadside impairment assessments, and among drivers whose behavior had gotten them arrested, only 30% to 49% would have counted as impaired under a 5 ng/mL rule. The foundation concluded that a per se THC limit cannot be scientifically supported, because unlike alcohol, THC levels in the blood do not track impairment in a reliable way.

Newer research explains why the line fails to identify someone who is too high to drive. In late 2025, researchers at UC San Diego’s Center for Medicinal Cannabis Research published a study in Clinical Chemistry that drew blood from 190 regular users after at least 48 hours of abstinence. Nearly a quarter still had 2 ng/mL or more; 5.3% were at or above 5, with a maximum of 16.2. Their simulated driving did not differ between groups. THC lingers in the blood of daily users long after its effects fade.

The opposite happens with occasional users. In occasional users, blood THC typically falls below 1 ng/mL within four to twelve hours of smoking, according to Norwegian Institute of Public Health researchers Hallvard Gjerde and Jørg Mørland, writing in Addiction, while studies measure impairment lasting up to several hours depending on the dose.

If the blood test can’t establish impairment, the police’s only other tool is the roadside assessment. In 2023 the same UC San Diego group put the standard field sobriety battery to a blind test in JAMA Psychiatry. Trained officers evaluated 184 regular users who had smoked THC or placebo. One hour after smoking, they judged 81% of the THC group impaired, and 49% of the placebo group. In one sense the test worked: officers correctly flagged most of the intoxicated group. But 19% of those under the influence were judged unimpaired, and 49% of the sober placebo group were judged impaired. The test is not good enough, and many people who are fit to drive will be found unfit on the strength of a test that cannot reliably tell them apart.

The lead author, Thomas Marcotte, called the tests useful additions to a driver evaluation but not accurate enough on their own to determine THC impairment.

Legally, little has changed when it comes to the tests. Massachusetts’ Supreme Judicial Court ruled in Commonwealth v. Gerhardt (2017) that in marijuana cases officers may describe what they observed but may not testify that a driver passed or failed the tests, which were built for alcohol. More recently, New Jersey’s Supreme Court in State v. Olenowski (2023) admitted drug-recognition expert testimony with limits on what the expert may claim, noting that no state has discontinued the program.

Feeling Sober Isn’t Being Sober

That leaves the driver’s own judgment, the standard many prosecutors fall back on. The evidence there is mixed in a way that matters. In a 2020 on-road trial published in JAMA, Thomas Arkell and colleagues had 26 occasional users vaporize cannabis and drive 100 kilometers of Dutch highway. Lane weaving rose by about 2.3 centimeters at 40 to 100 minutes, close to the trial’s benchmark for a 0.05% blood alcohol level, and most drivers felt the effect.

A UC San Diego trial of 191 regular users, published in JAMA Psychiatry in 2022, found the problem arrives later. Participants hesitated to drive right after smoking, but by 90 minutes about two-thirds said they were ready while their simulator scores had not improved; most had recovered by four and a half hours. The FDA reached a similar conclusion about sleeping pills in 2013, warning that next-morning alertness can be impaired in people who do not feel drowsy.

Adding to the difficulty of judging DUI on behavior, regular users consistently show less impairment than occasional ones.

With the rise of high-potency flower, rosin, BHO, edibles and other concentrates, the science on the drug’s effects has also sharpened. A 2021 meta-analysis of 80 studies led by Danielle McCartney at the University of Sydney found THC impairs lane control and divided attention for roughly three to ten hours, depending on dose, whether it is smoked or eaten, how demanding the task is and the user’s habits. Complex and unexpected tasks recover last.

What the Crash Data Shows

How much of what controlled experiments find shows up as real-world crashes is smaller and more contested than the public debate usually assumes. NHTSA’s case-control study in Virginia Beach, comparing crash-involved drivers with drivers at the same locations and times, found THC-positive drivers about 25% more likely to crash before adjustment and no more likely once age, sex, race and alcohol were accounted for. That study measured presence, not concentration. The same study put a 0.08 blood alcohol level at nearly four times the crash risk.

Also in 2016, a meta-analysis by Ole Rogeberg and Rune Elvik in Addiction, correcting earlier reviews, found a statistically significant increase of low-to-moderate magnitude, with odds ratios around 1.2 to 1.4. A 2018 correction, co-authored with Michael White after he identified data errors, nudged the pooled odds ratios slightly lower, to about 1.2 to 1.3, without changing the conclusion. Two Norwegian Institute of Public Health researchers pushed back in the same journal, arguing that most of the pooled studies measured traces of recent use rather than acute intoxication.

Most per se laws predate this evidence; Washington’s 5-nanogram limit, for instance, arrived with the state’s 2012 legalization initiative.

What has clearly grown is exposure. After Canada legalized in 2018, the share of moderately injured drivers at four British Columbia trauma centers with THC at or above the 2 ng/mL legal limit rose from 3.8% to 8.6%, and the share at 5 ng/mL or more from 1.1% to 3.5%. In a 2019 study, the same team found drivers at 5 ng/mL or more may face higher risk, though the increase was not statistically significant.

But again, THC in the blood does not show that a crash was caused by impairment. More drivers testing positive can mean more people use cannabis, not that more people drive impaired. It is a bit like counting more stars through a bigger telescope: the count goes up, but that alone doesn’t tell you the sky got more crowded.

Two Rulebooks: Benzodiazepines and Weed

Consider a patient who manages chronic pain with prescription oxycodone and is stopped for speeding in Washington. To convict, the state needs evidence of impairment, usually starting with a roadside assessment. If the patient fails it, they can be arrested and their blood tested, but the blood result alone isn’t enough to convict. A medical cannabis patient stopped on the same road faces a different calculation: if the blood test comes back at 5 ng/mL or more, the number can decide the case on its own.

That gap is hard to justify on safety grounds. Benzodiazepines such as Valium and Xanax are dangerous at the wheel. A 2011 meta-analysis in Drug Safety found users about 60% more likely to crash in case-control studies and about 80% more likely in cohort studies, with the odds rising nearly eightfold when combined with alcohol. That is a larger and more consistent signal than the cannabis estimates above.

Yet American law treats the two drugs very differently.

The first rulebook: benzodiazepines and other prescription drugs. Benzodiazepine cases are generally prosecuted on evidence of impairment, and several states with “any amount” laws exempt drugs taken as prescribed. According to the Governors Highway Safety Association, 16 states have zero-tolerance laws for one or more drugs and 5 have per se limits, but these mostly target illegal drugs, and some exclude prescription medications altogether. Illinois’ zero-tolerance clause covers only drugs in the body from the “unlawful use or consumption” of a controlled substance. Georgia provides that a person legally entitled to use a drug is not in violation unless “rendered incapable of driving safely” by it. Arizona, which bans driving with any amount of a listed drug, gives an affirmative defense to people using a medicine as prescribed. Pennsylvania’s “any amount” rule covers Schedule I drugs and Schedule II or III drugs “not medically prescribed”; benzodiazepines, which are Schedule IV, fall under its impairment provision instead.

The second rulebook: cannabis. According to GHSA, 18 states punish THC by presence or by number, whatever the driver’s actual condition. The remaining states, including California, New York, Oregon, Massachusetts and New Jersey, prosecute cannabis like any other drug, on evidence of impairment.

Three states have tried to bridge the gap for patients, each differently. Illinois wrote its own patient exemption into the 5-nanogram statute: registered patients can be convicted only if impaired. Michigan’s Supreme Court reached the same place in People v. Koon (2013), holding that the state’s medical marijuana act overrides its zero-tolerance law for registered patients. Arizona went halfway. In Dobson v. McClennen (2015), its Supreme Court refused to extend the prescription defense to medical cannabis, because patients hold a certification rather than a prescription. Instead it let cardholders raise a defense if they can prove, by a preponderance of the evidence, that their THC level was too low to cause impairment. The burden falls on the patient.

Everywhere else in the per se and zero-tolerance group, the patient sits on the wrong side of the line.

What Fair Would Require

The two failure modes point to the same rule. A per se limit charges patients on a number that cannot establish impairment. An unconditional exemption has the opposite problem. A rule that protects patients without protecting impaired drivers has to close both gaps.

The starting point is parity. A patient on medical cannabis should face the same legal standard as a patient on oxycodone or diazepam, where impairment is the offense and presence alone is not.

But until lawmakers and scientists settle on a reliable way to measure impairment, and that day may still be years away, the burden falls on us. The honest truth is that most people can’t tell where “a little high” ends and “too high to drive” begins.

So the simplest rule is the best one: if you’re stoned, don’t drive.

For daily medical patients, who often build tolerance, the question isn’t whether to drive but when. Impairment after using cannabis can last roughly three to ten hours depending on the dose, according to the McCartney meta-analysis, and tolerance varies from person to person. Waiting it out, skipping the wheel after a new or higher dose and knowing your state’s rule, laid out in the table above, are the closest thing patients have to protection until the law catches up.

For occasional users, whose bodies aren’t used to THC, the case is even clearer. Call a ride, wait it out, or stay where you are. Defending cannabis means defending the people around us too. A crash takes only a second, and it can upend a stranger’s life as easily as your own.

Responsible starts the moment you decide whether to pick up your keys.

This article is for informational purposes only and is not legal advice. Impaired-driving laws vary by state and change often.