IN SAN FRANCISCO, the law has gotten ahead of technology for once. This week, the city became the first in the nation to ban government agencies from using facial-recognition software. The impulse to control a nascent technology that could pose real risks to civil rights is laudable. But a moratorium might be a better strategy than outright prohibition to limit the dangers of an unregulated but potentially useful tool.
There is no need to imagine a dystopian surveillance state in which a government can follow its citizens everywhere they go simply by scanning their faces. That state already exists. China’s ubiquitous cameras and facial-recognition software allow the tracking of its Uighur population, in particular. Officials also hold face databases of criminals, the mentally ill, known drug users and even those who have lodged grievances with the government. Russia is expanding its use of similar products.
Right now, police departments across the United States are experimenting with techniques eerily similar to those overseas. Cities have the capacity, through strategically placed cameras, to locate anyone with a criminal record, or they are testing pilot programs working toward that goal. Authorities are setting low accuracy thresholds for matches, against the advice of software manufacturers, while running pixelated images and even sketches through the programs. Systems that studies have suggested perform poorly when identifying women and people of color are not audited for bias. Defendants are apprehended with the help of the technology and never find out — even in court. This is happening, in many cases, without the communities under watch knowing a thing.
Americans have the advantage of living in a democracy where citizens have some say in just how closely the government can keep an eye on them. Right now, when it comes to facial recognition, they are not getting that chance. There’s potential for a legal regime that realizes the legitimate advantages that recognition techniques can offer, from finding trafficking victims to apprehending suspects in serious crimes, while also guarding against its abuse. But finding that regime will take time.
San Francisco’s law should be welcomed as a way to grapple with facial-recognition technology before it becomes so entrenched that Americans regard face scans as routine and inevitable. The country should be figuring out when police should be allowed to turn facial recognition on, rather than when they should be forced to turn it off. But hitting stop without a plan for further study might miss the opportunity to throw the right switches. Call it a pause, not a ban.