ecreational marijuana use is now legal (or about to be) in 11 states and Canada. Driving under its influence is not. But if police suspect someone is driving high, they’ll have a difficult time proving it: Marijuana tests typically rely on urine samples, take days to analyze, and may only indicate that someone has used marijuana over the past days or even weeks, not whether they are actively impaired.

In response, several startups say they are building “pot breathalyzers” that will instead (eventually, they swear) deliver results about recent marijuana use, on the spot. The most obvious use for these devices is patrolling roads. But there’s another place where it could be very important to tell the difference between legal marijuana use last weekend and current impairment: your job.

While stoners like to debate whether driving high is actually dangerous (for the record: multiple metaanalyses of multiple studies say yes; one study conducted by the National Highway Traffic Safety Administration says not really), there’s no doubt that marijuana use is getting people fired.

The U.S. Department of Transportation requires drug testing for “safety-sensitive” positions like pilots, bus and truck drivers, and railroad operators. Federal government contractors and those receiving federal government grants are required to maintain a “drug free” work environment. And manufacturing facilities, warehouses, and wholesalers may also drug test for safety reasons. After a workplace accident, employers may use a positive drug test to argue that injuries or damage were caused by a worker’s impairment, rather than the company’s negligence.

Marijuana is the most commonly flagged substance in workplace drug tests, according to Quest Diagnostics, which sells testing services to employers. Nearly 3% of all screenings the company processed in 2018 tested positively for the drug (4.4% of screenings tested positive for any substance).

As marijuana becomes legal in more states, however, the usefulness of a test that determines whether someone has used marijuana days or weeks in the past is starting to look about as useful as a test that determines whether someone has used alcohol in the last week. “If you get caught with cocaine in your system because it’s illegal, I probably have the right to fire you,” says James Spigener, who consults with organizations on safety practices as a chief client officer at Dekra. “Having marijuana in your system, in many places, is not a violation of any law.”

Accommodating an employee’s recreational marijuana use is not required by law in any state, though some require accommodations for medical marijuana. Still, the legal grounds aren’t clear cut. In 2015, the Colorado Supreme Court sided with Dish Networks after it fired a customer service representative who used medical marijuana to treat painful muscle spasms caused by his quadriplegia. It argued that even though marijuana use was legal under state law, it is still prohibited by federal law. In a similar case in Massachusetts, the state supreme court sided with an employee using medical marijuana to treat Crohn’s Disease. (The court cited Massachusetts’ medical marijana law, which states that a user should not be denied “any right or privilege”).

Even if these methods do work, there’s an additional hiccup: it’s not clear that such breathalyzers can detect marijuana that has been ingested rather than smoked.

Many employers don’t want to fire people for using legal marijuana on the weekend. Jim Reidy, director at large for HR State Council of New Hampshire, says he’s noticed that many employers are dropping their marijuana screenings. “Employers are desperate to find skilled employees,” he says. When he asks employers why they dropped the test, “part of it is [marijuana] is becoming legal. Part of it is also that if you’re not in a safety-sensitive position, employers don’t care.” Companies that hire for safety-sensitive positions can’t give up drug testing so easily, and they don’t have a test that can distinguish between legal use over the weekend and impairment caused by the drug.

SannTek Labs, one startup working on a pot breathalyzer, says it has seen interest from companies that do drug testing as a service for a more specific marijuana test. “We’ve looked at it just as much as the police opportunity,” says Noah Debrincat, the company’s CEO and founder. “The reason it appears we are focused on police is there is only a limited amount of information I can get across in these interviews, and everyone seems interested in the police aspect.”

Detecting any level of marijuana on the breath would be evidence that someone has used the drug within a relatively recent time period, Debrincat says. But that’s not easy to do because the concentration of THC is extremely low in the breath, even in those who are clearly intoxicated. Debrincat says that while someone who is legally impaired by alcohol might have 208 ppm of ethanol in the breath, a similarly impairing dose of marijuana would result in only about 0.00001 ppm of the drug in breath. SannTek Labs aims to detect this small trace by enhancing a biochemical test with nanotechnology. Instead of testing the vapor of the breath like an alcohol breathalyzer does, it tests particles of fluid that make it into the breath from the airway lining (this is different than a saliva test, which involves fluid that comes from the mouth rather than the airway). The big question is: Does it work?

I ran SannTek’s explanation by Harry Ruda, the director of the Centre for Advanced Nanotechnology at the University of Toronto. He called it “surprising.” It is a far from proven approach.

SannTek says it has run its own experiments with 20 participants. Another pot breathalyzer startup that uses a different type of technology, called Hound Labs, has also run a trial with an undisclosed number of participants. Neither study has been published or peer-reviewed, though a small study from researchers at the University of California, San Francisco did suggest it was possible to detect THC on the breath after smoking marijuana.

Scientists at the University of Pittsburgh developed a prototype using a different technology called carbon nanotubes, which they describe as “tiny tubes of carbon, 100,000 times smaller than a human hair.” The material is expensive, but Alexandar Star, a professor of chemistry and bioengineering who worked on the project, told OneZero that the approach only requires a tiny amount of it — about $1 worth — to create a sensor.

Even if these methods do work, there’s an additional hiccup: it’s not clear that such breathalyzers can detect marijuana that has been ingested rather than smoked.

And technology is only the first hurdle to a workable marijuana breathalyzer. There is no legal standard for what counts as “impairment” when it comes to marijuana use, and there is some evidence that the level of marijuana and level of impairment might not be perfectly correlated. “For alcohol, the law has declared that .08% [blood alcohol concentration] is the level in which you’re intoxicated,” says Spigener. “We don’t have that with marijuana. Nobody knows how much it impairs you or not.”

In other words, even if you can detect recent marijuana use, how do you prove that the level at which you’ve detected it causes impairment or not? How can you prove that trace amounts detected from previous use by the current tests aren’t causing some impairment? Without a legal limit, and research to back it up, there’s no easy answer.

Spigener favors a much simpler solution than the breathalyzer test. Instead of detecting substances, he wants to see companies directly test impairment — say, by asking pilots to type the same sequence of numbers as fast as they can before they fly, and noticing when their fastest time is relatively slow. These types of impairment tests are already used in conjunction with drug tests (the strategy that SannTek’s Debrincat favors). By not focusing solely on drugs, Spigener believes, companies might catch other good reasons that employees should not, say, operate heavy machinery. “What you may find out is the guy had a kid who was up all night and couldn’t get to bed,” he says. “No foul no harm. Go home and take a nap.”