An impressive trove of public records obtained by BuzzFeed shows just how pervasive facial recognition tech is. Law enforcement agencies are embracing the tech, often with a minimum of accountability or oversight. That's how toxic tech purveyors like Clearview -- whose software relies on a multi-billion image database scraped from the web -- get their foot in the door to secure government contracts.
Despite being used for years, facial recognition tech has yet to prove it's capable of recognizing the right faces more often than the wrong ones. The accuracy gets even worse when it's deployed to recognize faces of women and minorities -- and given law enforcement's history of disproportionate enforcement -- it will be minorities harmed by the inaccurate tech more often than not.
What BuzzFeed has done with these Clearview records is compile a searchable database that allows readers to see if their local agencies have tried out the tech. Clearview's tech has yet to be subjected to outside review and its method of obtaining images -- scraping them from public posts on the web -- leaves a lot to be desired in terms of accuracy. (Unfortunately, as the EFF's Dave Maas points out, this doesn't mean BuzzFeed has made the dataset public -- only its interpretation of the data. But we'll take what we can get.)
The upshot? Lots and lots of experimentation. The downside? Very little oversight or explicit permission. According to the information BuzzFeed obtained, more than 335 US law enforcement agencies have at least tried out Clearview's facial recognition AI, and many of those searches had nothing to do with investigations.
Several of the responding agencies appear to be paying little attention to the actions of their employees:
Officials at 34 of those organizations said they were unaware that their employees had signed up for free trials until our questions prompted them to look.
Meanwhile, others pretended they had no responsive documents until asked twice:
Officials at another 69 entities at first denied their employees had used Clearview but later determined that some of them had.
While a tally of 335 law enforcement agencies may seem minute in comparison to the total number of law enforcement agencies in the United States, it would be wise to remember this dataset is far from complete. Nearly 100 agencies refused to answer definitively whether or not their employees had explored Clearview's offerings. Nearly 1,200 agencies have yet to turn over requested documents.
Meanwhile, Clearview has been out there touting law enforcement successes few law enforcement agencies will acknowledge. It also encourages the perception it is currently partnering with hundreds of cop shops while trying to divert attention away from its willingness to grant access to anyone interested in its unproven tech, including government agencies in countries known for their human rights abuses.
This data shows there's more interest in Clearview than law enforcement (as a whole) is willing to admit publicly. It also shows cops are playing with unproven tech -- quite possibly using images of people who aren't suspected of anything -- without the knowledge of their supervisors or government officials charged with overseeing their actions. What's left unsaid -- or unresponded to -- is at least as concerning as what has been admitted publicly. There's a rogue AI on the loose that links cops to a database filled with billions of images scraped from websites without their -- or their end users' -- approval.
Clearview is entirely problematic. But these agencies' willingness to exploit and examine this tech is even more so, considering the damage to rights and civil liberties the careless use of unproven facial recognition tech can cause.
With thanks to our Friends over at : Techdirt.
Read the Full Story : Click here