Facebook is ending its auto-tagging facial recognition feature — for now

A significant move, but not one that really constitutes a promise for the future.

Westend61/Westend61/Getty Images

Facebook is completely turning off its facial recognition systems, parent company Meta announced today in a blog post. More than a billion users’ facial recognition data will be deleted from Facebook’s servers as part of the shutdown.

The Face Recognition system has been used for more than a decade on the social media platform, both for tagging users in photos and for creating automatic alt text for other images on the site. The feature was game-changing at the time of its introduction, as anyone who was on Facebook then will remember; rather than forcing users to tag photos one by one, they could now easily review automatic tags that Facebook had identified.

After first reviewing the ways in which facial recognition technology can, indeed, be useful, Jerome Pesenti, VP of Artificial Intelligence, writes: “There are many concerns about the place of facial recognition technology in society, and regulators are still in the process of providing a clear set of rules governing its use.” This, he says, is why Facebook will no longer use facial recognition software — there’s just too much uncertainty about its harms. But Facebook is hedging its bets here anyway.

This isn’t forever — Pesenti is very careful with his words. It’s obvious from the get-go that he (and Facebook) want this announcement to be seen as a grand gesture. “This change will represent one of the largest shifts in facial recognition usage in the technology’s history,” he writes. Talk about self-aggrandizing.

Notably, Facebook’s decision here is not all-encompassing. Yes, a billion users’ faces will be deleted from the system, but the system itself will still very much exist. DeepFace, the algorithms that power Facebook’s facial recognition system, isn’t going anywhere, The New York Times confirms.

And Facebook hasn’t committed to never using facial recognition in the future — just at this particular moment. A year from now, for example, Mark Zuckerberg’s metaverse might use your facial data to recommend a pair of digital shoes. As recently as February, Andrew Bosworth said the company was considering adding facial recognition to its smart glasses. There are no long-term promises here.

A turning point — Despite Facebook’s attempts to toe the line here, the company’s decision to step away from facial recognition — even temporarily — is, indeed, significant, both for the social network and for facial recognition technology as a whole.

Facebook’s house-made facial recognition technology has proven thorny for the company over the years. In January 2020, the company settled an Illinois lawsuit for $550 million over claims Facebook’s facial recognition software violated local biometric data laws. Many national advocacy groups have raised concerns over Facebook’s banks of facial data, and the Electronic Privacy Information Center even filed two complaints with the FTC over the technology.

Public opinion of facial recognition technology has shifted significantly since Facebook first introduced its auto-tagging feature in 2010. Various local governments have banned the technology in recent years, and huge tech companies like IBM and Microsoft have put moratoriums on their own research into facial recognition AI.

As the most popular social network in the world, Facebook’s decision here carries some weight. The tide of facial recognition still moves forward with a rapid pace some might call alarming, though. On the other side of the equation, startups like Clearview AI are scraping the internet for any and every face that’s ever been uploaded. And law enforcement is still adopting the technology — more with every passing day.

Facebook’s decision to wait this one out until more comprehensive federal policies have been enacted isn’t exactly altruistic, but it is savvy. Rather than choose a side, the social network has decided to sit this one out until someone with more power decides what’s what.