A patent reveals the rideshare app could predict if users are in an "unusual" state.

Though Uber revolutionized the taxi industry in general, as anyone who likes to pair a couple bottles of wine with dinner will tell you, digital taxi apps have also revolutionized the extent to which people can enjoy a night out. Of course, as designated drivers have known for decades, shepherding around drunks can be the exact opposite of enjoyable, so it’s easy to see why companies like Uber might want to be able to tip their drivers off to intoxicated customers. However, one idea Uber has for checking users’ sobriety might leave you thinking twice about how you use their app.

According to a U.S. patent application filed in 2016, but apparently made public on June 7, Uber is interested in the possibility of “predicting user state using machine learning,” as the patent is titled, and possibly alter your service if the company is worried about your condition. More specifically, the application says that “incidents” can occasionally occur when users “behave uncharacteristically,” so this proposed technology would look for signs of such uncharacteristic app usage like typos, imprecise pressing of links and buttons, and even things like how the phone is being held or the speed the user is walking. Other risk factors like time and location could also be taken into account. Interestingly, words like “alcohol,” “drugs,” “drunk,” or “intoxicated” appear not to be used anywhere in the application—but identifying these “states” would seem to be the direction this technology is headed.

As the Telegraph points out, and the language of the patent confirms, such behavioral learning could be applied to and potentially benefit both users and drivers. For instance, potential benefits to riders might include “modifying pickup or dropoff locations to areas that are well lit and easy to access,” as the patent states, or specifically matching these riders with “providers with experience or training with users having an unusual state.”

However, some experts worried that such information could be used to strand, or even take advantage of, intoxicated customers. “It could lead to the possibility of some drivers avoiding drunk passengers and in the worst cases ‘drunk hunting,’” Andre Spicer, a professor from Cass Business School in London, told the Telegraph. He later added another annoying possibility: “It could mean that they offer different pricey for drunk passengers, which means Uber could take advantage of a user’s state to charge them more.”

Those concerns are just the tip of the iceberg, too. Attempting to predict intoxicated states could become a privacy issue, especially if Uber is essentially logging a dossier on how often it thinks people are getting trashed, info that could potentially be harmful to users if revealed down the line. Meanwhile, if drunk people are denied rides and instead choose to drive, that’s a different kind of safety concern.

As with all patents, it’s important to remember that a patent is just a concept, and big tech giants like Uber often file patents simply to protect an idea, even without the intention of implementing it. Still, as far as concepts go, this one is certainly an doozy—and possibly an unsettling glimpse into just how much info our smartphone might be able to uncover about our behavior… whether it hurts your chances at getting a cab or not.