By: George Daoud
On its surface, facial recognition software does not seem threatening. After all, many people use it to unlock their latest iPhone. In fact, in the hands of law enforcement and federal agencies, it can be used to track and apprehend criminal suspects or victims of crimes. Companies like Clearview AI, for example, produce facial recognition software that they claim is created for just that purpose.
The Clearview AI concept is relatively simple. They provide law enforcement with a tool to upload someone’s picture onto their software, then, the algorithm does the rest. Through online scraping, the algorithm searches Facebook, Twitter, Google, Instagram, LinkedIn, and endless other web sources, to identify a host of information about the selected person within seconds. This information can include address, place of work, friends, recent travel, and arrest records. Needless to say, the usefulness of this type of technology is apparent for certain law enforcement and surveillance purposes, but the question is where to draw the line in terms of privacy.
To further this privacy dilemma, Clearview AI’s entire client list was recently exposed in a massive data breach. This discovery shed light on the fact that Clearview had not only sold its technology to more than 2,200 law enforcement organizations, government agencies, and security companies around the world, but that some of these entities were not law enforcement at all. Well known companies like, Macy’s, Kohl’s, Walmart, and the NBA to name a few, were amongst the Clearview clients that were found to be regularly using the software. Furthermore, it was discovered that Clearview’s founders gave potential investors and friends free access to the technology, escalating the public unease that already existed.
How Does It Work?
Facial recognition software is a type of artificial intelligence (AI) that uses ‘machine learning’ and more specifically ‘deep learning’ methods to operate. Deep learning in AI involves digital algorithms that function similarly to a human brain, and analyze information using artificial neural networks. These neural networks are composed of interconnected neurons, like a brain, that break down and analyze data. Moreover, each neural network is composed of layers, namely an input, hidden, and output layer. The hidden layer is where an AI algorithm does the majority of its computing. In order for facial recognition software to function at its maximum potential, it requires a multitude of hidden layers which require more data points to provide greater accuracy and more efficient results. In these hidden layers, however, is where companies, like Clearview AI, focus their development, and where the public concern for privacy protections lies.
Since the hidden layers of facial recognition algorithms require more data to be accurate, everyday individuals, and those wrongly suspected, could be subject to massive invasions of privacy through digital data mining. This not only brings the possibility of Fourth Amendment unreasonable searches and seizures violations into the forefront, but also highlights the level of scrutiny that should be given to personal digital information in the eyes of the law.
In the 2018 Supreme Court case, Carpenter v. U.S., the Court decided that the warrantless acquisition of a person’s cell-cite information by law enforcement was an unconstitutional search in violation of the Fourth Amendment. Furthermore, the Court made the distinction between when privacy should and should not be expected. They deemed that digital cell-cite information given to a third party, like a phone company, and acquired by police, was only pseudo-voluntary. Therefore, the owner of the information does not have a genuine choice in terms of allowing third-party access. This pseudo-voluntariness, according to the Court, leads a reasonable person to have a higher expectation of privacy for their information. If a reasonable person does not think that their digital information would be subject to regular scrutiny by a third-party, then their privacy expectation is inviolable by law enforcement. That being said, where does that leave a company like Clearview AI?
More Questions Than Answers…
The crux of the problem with current facial recognition software, is that, unlike in Carpenter, the information that is utilized by law enforcement will not likely hold the same privacy weight in a court of law. This is mainly due to the fact that the information used, is publicly available and is voluntarily provided to third parties. When one creates an online profile and purposefully puts their digital footprint on the web, the expectation for privacy is assumed to be lessened. As a result, the liability that companies and law enforcement agencies will face, when using facial recognition to track people, will be drastically reduced. This has the potential to not only increase its unfettered usage in both the private and public sector, but also to incentivize future developers to push the boundaries. How much and what type of information can be mined and utilized by an algorithm’s hidden layers are open to interpretation. The only concern companies have right now is getting the most accurate results. The question is though, who watches the watchmen?
There is no doubt that there are benefits to facial recognition software, especially in the context of law enforcement. However, when private companies and individuals can peek into the lives of anyone through a picture, protection at the cost of privacy may not be as good as it sounds.
Kashmir Hill, The Secretive Company That Might End Privacy as We Know It, The New York Times (Jan. 18, 2020), https://www.nytimes.com/2020/01/18/technology/clearview-privacy-facial-recognition.html.
Ben Gilbert, Clearview AI Scraped Billions of Photos from Social Media to Build a Facial Recognition App That Can ID Anyone — Here’s Everything You Need to Know About the Mysterious Company, Business Insider(Mar. 6, 2020),https://www.businessinsider.com/what-is-clearview-ai-controversial-facial-recognition-startup-2020-3#then-in-march-another-new-york-times-piece-on-the-company-revealed-another-stunning-detail-the-companys-founders-casually-gave-access-to-the-software-to-potential-investors-and-friends-who-immediately-abused-it-6.
Peter Margulies, Surveillance By Algorithm: The NSA, Computerized Intelligence Collection, and Human Rights, Fla. L. Rev. (July 2016), https://scholarship.law.ufl.edu/cgi/viewcontent.cgi?article=1321&context=flr.
Carpenter v. United States, 138 S. Ct. 2206 (2018).
UNIDIR, The Weaponization of Increasingly Autonomous Technologies: Artificial Intelligence a Primer for CCW Delegates, U.N. Docs.No. 8 (2018).