A very questionable facial recognition tool being offered to law enforcement was recently exposed by Kashmir Hill for the New York Times. Clearview — created by a developer previously best known for an app that let people put Trump’s “hair” on their own photos — is being pitched to law enforcement agencies as a better AI solution for all their “who TF is this guy” problems.
Clearview doesn’t limit itself to law enforcement databases — ones (partially) filled with known criminals and arrestees. Instead of using known quantities, Clearview scrapes the internet for people’s photos. With the click of an app button, officers are connected to Clearview’s stash of 3 billion photos pulled from public feeds on Twitter, LinkedIn, and Facebook.
Most of the scrapees have already objected to being scraped. While this may violate terms of service, it’s not completely settled that scraping content from public feeds is actually illegal. However, peeved companies can attempt to shut off their firehoses, which is what Twitter is in the process of doing.
Clearview has made some bold statements about its effectiveness — statements that haven’t been independently confirmed. Clearview did not submit its software to NIST’s recent roundup of facial recognition AI, but it most likely would not have fared well. Even more established software performed poorly, misidentifying minorities almost 100 times more often than it did white males.
The company claims it finds matches 75% of the time. That doesn’t actually mean it finds the right person 75% of the time. It only means the software finds someone that matches submitted photos three-quarters of the time. Clearview has provided no stats on its false positive rate. That hasn’t stopped it from lying about its software and its use by law enforcement agencies.
A BuzzFeed report based on public records requests and conversations with the law enforcement agencies says the company’s sales pitches are about 75% bullshit.
Clearview AI, a facial recognition company that says it’s amassed a database of billions of photos, has a fantastic selling point it offers up to police departments nationwide: It cracked a case of alleged terrorism in a New York City subway station last August in a matter of seconds. “How a Terrorism Suspect Was Instantly Identified With Clearview,” read the subject line of a November email sent to law enforcement agencies across all 50 states through a crime alert service, suggesting its technology was integral to the arrest.
Here’s what the NYPD had to say about Clearview’s claims in its marketing materials:
“The NYPD did not use Clearview technology to identify the suspect in the August 16th rice cooker incident,” a department spokesperson told BuzzFeed News. “The NYPD identified the suspect using the Department’s facial recognition practice where a still image from a surveillance video was compared to a pool of lawfully possessed arrest photos.”
The NYPD also said it had no “institutional relationship” with Clearview, contradicting the company’s sales pitch insinuations. The NYPD was not alone in its rejection of Clearview’s claims.
Clearview also claimed to be instrumental in apprehending a suspect wanted for assault. In reality, the suspect turned himself in to the NYPD. The PD again pointed out Clearview played no role in this investigation. It also had nothing to do with solving a subway groping case (the tip that resulted in an arrest was provided to the NYPD by the Guardian Angels) or an alleged “40 cold cases solved” by the NYPD.
The company says it is “working with” over 600 police departments. But BuzzFeed’s investigation has uncovered at least two cases where “working with” simply meant submitting a lead to a PD tip line. Most likely, this is only the tip of the iceberg. As more requested documents roll in, there’s a very good chance this “working with” BS won’t just be a two-off.
Clearview’s background appears to be as shady as its public claims. In addition to its founder’s links to far right groups (first uncovered by Kashmir Hill), its founder pumped up the company’s reputation by deploying a bunch of sock puppets.
Ton-That set up fake LinkedIn profiles to run ads about Clearview, boasting that police officers could search over 1 billion faces in less than a second.
These are definitely not the ethics you want to see from a company pitching dubious facial recognition software to law enforcement agencies. Some agencies may perform enough due diligence to move forward with a more trustworthy company, but others will be impressed with the lower cost and the massive amount of photos in Clearview’s database and move forward with unproven software created by a company that appears to be willing to exaggerate its ability to help cops catch crooks.
If it can’t tell the truth about its contribution to law enforcement agencies, it’s probably not telling the truth about the software’s effectiveness. If cops buy into Clearview’s PR pitches, the collateral damage will be innocent people’s freedom.