As with so many other fields, 2020 bore witness to a dizzying series of developments within the facial recognition policy sphere. At ACT| the App Association, we did our best to stay abreast of it all. We compiled a taxonomy of legislative responses to facial recognition and created an infographic for facial recognition use-cases to illustrate the diversity of applications for a technology often viewed one-dimensionally.
Even as facial recognition retained its place in the spotlight last year, the long-term implications of the technology remain muddled. While there were moments of clarity regarding its risks, some unexpected use-cases contradicted the predominant narrative and prompted even some skeptics to take a moment of pause. In the clarity bucket, the world finally received evidence of a harm that activists and academics have long warned us about. In August, the New York Times reported the first documented wrongful arrest of an individual, Robert Williams, a black man, on the basis of a faulty identification by facial recognition technology and the officers working the case. Revelations of similar injustices followed Williams’s plight, emphasizing the downside risk of continued unregulated law enforcement use of facial recognition and underscoring that the sharing of this burden will not be equal.
As for contradictions, a surprising constellation of law enforcement, protestors, and onlookers at last year’s protests and riots reached for facial recognition as a tool. On the one hand, many worried that law enforcement’s widely-reported usage of facial recognition to identify and track protestors at peaceful Black Lives Matter protests could lead to a chilling of protected First Amendment expression. On the other, the tables were turned somewhat when some activists created their own facial recognition algorithm and database to identify police officers who concealed their badge numbers. Already in 2021, police and internet users forged a surprising alliance in the aftermath of the Capitol insurrection last month. As law enforcement continues to seek the identity of individuals who stormed the Capitol, they, along with some help from internet users, rely on facial recognition to create leads from footage the protestors and others uploaded during the day’s events.
Finally, last year brought to the fore a number of policy trends that remain far from resolved and will be important to continue to track in 2021. Follow along as we discuss these in greater detail.
FTC Taking an Aggressive Stance?
As the country greets a new administration with new leadership at regulatory bodies throughout the federal government, it’s worth asking if we are to see a new paradigm vis-a-vis facial recognition. Those reading the tea leaves got a glimpse of what that could look like on January 11, when the Federal Trade Commission (FTC) voted 5-0 to settle with the photo storage app Everalbum over allegations of deceptive facial recognition policies. Key to the case were allegations that Everalbum misrepresented consumers’ ability to opt-out of the company’s facial recognition features and that the company falsely claimed that it would delete photos and videos once consumers deactivated their accounts. While the FTC’s pursuit of an actor it views as contravening its own privacy policy is commonplace, what was new about the case was the punishment that the FTC doled out. FTC’s settlement with Everalbum requires the company to delete any facial recognition models or algorithms developed with Everalbum users’ photos or videos. While it is certainly relevant that the FTC did not have the ability to monetarily punish Everalbum in this case due to statutory constraints, it is still huge news that the FTC was (unanimously, no less) willing to force a company to unwind its entire algorithm because some of the underlying data was collected unlawfully. Not only does this signal that the FTC will likely take a tougher stance against facial recognition purveyors generally going forward, but it also could have enormous ramifications across the entire data economy. While plenty of companies have been docked by the FTC for similar misrepresentations within their privacy policies, imagine if Facebook, for example, was forced to forfeit their entire timeline algorithm for a similar infraction. Subsequent FTC action in other spaces could be an absolute gamechanger, and surely ups the ante for companies to ensure that their privacy policy properly reflects their business model and data transactions, external and internal.
It’s also worth noting that the FTC’s action in this case culminated several months-worth of rhetoric from the two Democratic commissioners that telegraphed facial recognition’s position on the enforcement radar. Cementing this, Commissioner Chopra wrote in statement accompanying the complaint that “facial recognition technology is fundamentally flawed and reinforces harmful biases,” and that he “supports efforts to enact moratoria or otherwise severely restrict its use.” The two Democrats will presumably soon be joined with a third to form a majority for at least the next four years. Interestingly though, none of those Democrats will be (now ex-) Commissioner Chopra, long seen as a potential chairman under a Democratic administration. Chopra was instead nominated to head the Consumer Financial Protection Bureau (CFPB), and while he will surely delve into technology issues there, he’ll have far less sway over the facial recognition industry writ-large than if he had remained at the FTC. Meanwhile, Commissioner Rebecca Kelly Slaughter, the other Democrat currently serving, was recently designated as acting chair. Slaughter has been also critical of algorithmic bias and certain facial recognition technologies, though it remains to be seen if she’ll be as strident as Chopra. Nevertheless, the harsh Everalbum settlement gives us good evidence that facial recognition is in for enhanced scrutiny under the Biden FTC.
Legislation from Above?
In the legislative branch, unified Democratic control of the House and Senate could potentially create the opportunity for a breakthrough among the growing number of members expressing a desire to regulate facial recognition technology. Indeed, support for facial recognition regulation could potentially go as high as the Oval Office. As a presidential candidate, Vice President Harris touted reform of facial recognition technology as part of her criminal justice platform. Meanwhile, while President Biden has not spoken directly to the topic, he voiced past support for criminal justice reform broadly. Any large criminal justice package would be the most likely vehicle for facial recognition legislation.
So, what’s on the table? As we detailed early last year, Members of Congress introduced several bills, some bipartisan, some not, that ran the gamut of facial recognition issues – from commercial uses to “ongoing” (72 hours or longer) federal surveillance. However, the most stringent bill of 2020 came later in the year when Rep. Pramila Jayapal (D-WA) unveiled legislation that would outright ban the use of facial recognition technologies by federal agencies. As we look to 2021, the House Oversight & Reform Committee (upon which several of Jayapal’s co-sponsors sit) will likely play a leading role. Back in 2019, members of the House Oversight & Reform Committee held a series of hearings on the topic and were rumored to have a legislative vehicle in the works, though those efforts derailed following the death of chairman and issue champion Rep. Elijah Cummings (D-MD). With the commencement of the 117th Congress, all of the bills introduced last year are now null and will need to be reintroduced. Look for House Oversight & Reform to jumpstart the process.
Legislation from Below?
Finally, activity at the state and local level is another important vector to monitor. As we wrote back in March 2020, state action on facial recognition is far exceeding that of the federal government. That continues to be the case as, through the rest of 2020, legislative activity continued at a rapid pace. In Vermont, Governor Phil Scott signed into a law the first statewide moratoria on facial recognition technology. Massachusetts legislators passed a similar bill weeks earlier, though Governor Charlie Baker has so far refused to sign it. That effort built upon local bans of the technology in Cambridge and Boston. New York City has so far instituted a more limited moratorium on face recognition in schools until 2022, while a broader ban remains on the table. Portland, Oregon, has perhaps gone the farthest of any locality, banning both private and public uses of the technology. Into 2021, we expect exploration of similar measures to continue, and it is possible that state or local level legislation could serve as the template, prompt or both for federal action. The App Association will remain diligent in tracking these efforts through the legislative process.