A Stalker's Dream: Startup Scraped Billions of Images From Social Media for Facial Recognition AI
Compounding the concerns is the relative opacity surrounding the company and its Australian-born founder and CEO Hoan Ton-That, a self-taught app developer, engineer, and entrepreneur whose imprint on the tech ecosphere has been relatively limited until now.
According to information gleaned from the Times (we were unable to locate any patents for the company's technology), Clearview's system, which has apparently been licensed by hundreds of law enforcement agencies and a "handful of companies for security purposes", is relatively straightforward. Officials – who can try the technology free for 30 days – upload a pic to Clearview's website or app. Its AI then returns matches from a database that was built by scraping over three billion images from websites and apps including Venmo, Facebook, and YouTube. In a FAQ obtained by the Times, the company explains:
"When you upload a photo to Clearview ,our software analyzes the hundreds of features that make up the face and search for a matching face in our image database. Whether the software finds no results, similar results, or possible results is determined by how closely the facial features match those of another face in our database . The certain that the uploaded face and search results match, the lower the delta number will be. You can find the delta number by hovering over the word 'possible' or 'similar' in the search results."
The same document claims that "Clearview has the most accurate facial identification software in the world, with a 98.6% accuracy rate" and that although the company had a 30-60% hit rate, they were "adding hundreds of millions of new faces every month and expect to get to 80% by the end of 2019."
While none of this is terribly surprising, what is troubling is that the app's programming contains code that could turn essentially arm anyone with the technology. Hill reports that "the computer code underlying its app... includes programming language to pair it with augmented-reality glasses; users would potentially be able to identify every person they saw. The tool could identify activists at a protest or an attractive stranger on the subway, revealing not just their names but where they lived, what they did and whom they knew."
I've had two demos of Clearview & the results were frightening/stunning in their accuracy. Both demonstrations involved me giving blurry screenshots of a video & Clearview was able to identify both people (friends who had consented) even though they barely have a presence online https://t.co/53YULsJWj6— Yashar Ali ? (@yashar) January 18, 2020
While a memo provided by Clearview to potential clients outlines the legality of the technology as it relates to law enforcement (it positions Clearview "as a search engine of publicly available images," that "pulls and compiles publicly available images from across the Internet into a proprietary image database to be used in combination with Clearview's facial recognition technology," which makes it seem more like Yandex than Black Mirror), it does not address the potential for abuse. As the Times reports, both police and Clearview's investors believe access to the technology is likely to made available to the public at some point in the future. When asked about this by Hill, Ton-That acknowledged that his invention could very well open the door to stalkers, saying that "there's always going to be a community of bad people who will misuse it."
And of course he's correct. When asked about the company's misuse of Facebook's image repository, Ton-That glibly replied that "a lot of people are doing it... Facebook knows." A Facebook spokesperson responding to the Times said that the company was reviewing Clearview's use of the site and that they would "take appropriate action if we find they are violating our rules." It is unclear, however, what that action might - or even could - be. Ton-That does claim, however, that because Clearview's technology is only capable of scraping publicly available images, people who change their social media settings to disallow indexing by Google might slip under their radar.
While these concerns may remain a minor inconvenience for Ton-That and his investors who are enjoying the lack of regulation on facial recognition technology in the United States, it's a reality that other countries are taking seriously. The European Commission, taking its cues from the totalitarian nightmare that China has created, is considering a three to five year ban on public use of facial recognition technology throughout the EU so that sensible legislation that respects human dignity and privacy while attending to the needs of national and global security might be formulated.