Why Doesn’t the Public Trust Facial Recognition?
Does the track record of some firms who capture peoples’ facial images without their consent make it impossible for facial recognition to get a fair hearing in the court of public opinion – or are there deeper issues concerning the technology?
Facial recognition company Clearview AI is by no means a stranger to controversy, but May 2022 must go down as a bad month in the courts even by their standards.
At the beginning of the month the company agreed that it would not sell its facial recognition technology to most private firms in the US in a settlement that was reached at a federal court in Illinois.
The company had previously come under the spotlight in 2020 when its database containing billions of faces was breached. The faces were scraped from social media and, at the time, privacy advocates in the US condemned the company.
Also in 2020, the American Civil Liberties Union (ACLU) and the law firm Edelson PC filed a lawsuit against Clearview AI stating that it had taken all the images without peoples’ consent. The ACLU said that this was in violation of Illinois’ Biometric Information Privacy Act, BIPA 1. Prior to the lawsuit, buyers of the technology included the Chicago Police Department and the office of the Illinois Secretary of State.
A few weeks after the findings of the Illinois court, Clearview AI was in trouble again when it was fined more than £7.5 million by the UK’s privacy watchdog and told to delete the data, and to stop obtaining and using the personal data, of UK residents.
The UK Information Commissioner’s Office (ICO) said that the company’s practice of gathering images from the internet to create a global facial recognition database breaches UK data protection laws.
Clearview AI chief executive Hoan Ton-That said: ‘I am disheartened by the misinterpretation of Clearview AI’s technology to society.’
In this one sentence Clearview’s CEO has distilled the diametrically opposed positions taken by champions of facial biometric technologies and privacy advocates: utility v. privacy. Because it wasn’t so long ago (August 2020 in fact) that Clearview AI was seen useful enough to the US Department of Homeland Security (DHS) to award them a contract to give Immigration and Customs Enforcement (ICE) access to the facial recognition firm’s technology.
In an article in Nextgov 2, the information resource for US federal technology decision makers, Rob Mungovan explores this dichotomy between utility v. privacy and looks at the controversies and positive use cases that make this such a battleground for public argument.
Regarding privacy concerns about data security, the article acknowledges that personal data breaches continue to rise in volume and severity around the globe, but that new, innovative approaches are evolving all the time to help mitigate this vulnerability. One example is the concept of decentralised biometrics, which makes it virtually impossible for a hacker to access full facial datasets.
Many people do feel that sharing their personally identifiable information (including facial data) with the state feels intrusive. But the state has been collecting sensitive PII since records began and, as Clearview AI’s admittedly over-zealous approach to data scraping shows, people are willing to share all manner of data with commercial entities like Snapchat, Facebook and TikTok.
On the question of machine-learnt bias, the article claims that with a 99% accuracy rate of distinguishing individuals within minority groups or ethnicities, facial recognition technologies are so accurate that, rather than being a source of bias, facial recognition capabilities actually help overcome bias at the human level.
To a large extent, the debate between facial biometric technologists and privacy advocates centres around facial recognition for law-and-order purposes, where a pool of facial image data might be searched with the intention of matching one of the faces to a watch list – called one-to-many (1:N) searches.
In most use cases, the ID and secure document industry uses facial recognition to compare a person’s face (either present or not) to a face presented at the time of enrollment for access to a system or a service – so called one-to-one (1:1) matching.
The actions of some technology companies don’t help to increase the public trust in facial biometrics. The argument that what Clearview AI does is facial recognition (or facial surveillance when applied to larger group settings) and what our industry does is facial authentication, is not one that gives confidence to public concerns over privacy and data security.
What is important in this debate is continuing to stress the benefits of facial authentication for government and commercial applications – safer and more convenient travel, providing fast access to public services, support for people in crisis situations, enhancing person-not- present services...these are the things that will tip public opinion into accepting facial biometrics.
Subscriber content
Read the full article
Full access to ID & Secure Document News articles, newsletters and archives.
