During the height of the COVID-19 pandemic, when mask mandates became commonplace both in public and in private, tech vendors began selling products they claimed could detect whether someone was wearing a mask — or not. With press releases and flashy demonstrations, the vendors attracted the attention of critics skeptical about the solutions’ capabilities and potential surveillance applications. Allied Market Research optimistically predicted that the market would be worth over $1 billion by 2027.
Now, as mask mandates lift in countries around the world — if prematurely, according to some health experts — the dust is beginning to settle. While the demand for mask detection technologies is steadily declining, the products have had far-reaching effects with implications for privacy and security, interviews with vendors suggest.
For example, Shaun Moore, the CEO of Trueface, says he doesn’t see customers that have already purchased Trueface’s mask detection technology winding down their usage anytime soon. Like many of the vendors delivering mask detection as a service, Trueface, which was acquired by biometric security company Pangiam in 2021, came from a facial recognition background. The company got its start applying algorithms to camera footage to extract an abundance of data, including license plate recognition and object detection, before broadening its focus to biometrics.
“We started developing both mask detection and the ability to do face recognition with a mask around April of 2020, following the outbreak of COVID-19 in the U.S. Roughly 50% of our customers asked that we update our software with mask detection so they could programmatically tell their customers to pull down their mask,” Moore told TC via email. “We plan to keep mask detection and recognition with a mask on as part of our [product] in the event they are needed again.”
As regular readers of this site are well aware, facial recognition is a flash point for controversy. While companies like Trueface claim that they engage only in “responsible” deployments of the technology, recent history is filled with examples of facial recognition abuse, such as software developed by Huawei and others to recognize members of the targeted Uyghur minority group. Numerous studies including the landmark Gender Shades project have also shown facial recognition technologies to be susceptible to various biases, including gender, racial, and ethnic biases. Police have made several wrongful arrests of Black suspects based on faulty facial recognition evidence.
Trueface declined to name which customers are currently using its mask detection and facial recognition products, but the company previously won a contract from the U.S. Air Force to “secure base access and safety.”
Motorola Solutions, another vendor that began providing mask detection products during the pandemic, says that any customer that purchased its Avigilon Control Center 7 (ACC7) video management software and the necessary hardware can still access its mask detection technology free of charge. (ACC7 is maintained by Avigilon, a Canadian surveillance camera company that Motorola Solutions acquired in March 2018.) Motorola Solutions didn’t supply a customer list when asked, but according to NBC, Avigilon at one point had contracts with school districts, police departments, and housing authorities in the U.S.
“[Our] ‘No Face Mask Detection’ technology is a video-based detection technology that is able to … detect objects in the camera’s field of view, classify them as humans, and determine whether the subject is not wearing a mask,” Motorola Solutions spokesperson Elizabeth Skube told TC in an email. “In addition to alerts for security operators, users can generate enterprise-wide reports with statistical analysis over time to help employers address concerns … For now, the feature will remain available for customers to turn off or continue to use at their discretion.”
Like Motorola Solutions, Rhombus Systems, a security system supplier headquartered in Sacramento, California, began including face mask detection as a part of its standard platform several months ago (in January 2021). Companies can use it to receive alerts via push notification or email whenever the system detects that someone isn’t wearing a mask, CEO Garrett Larsson told TC via email.
“We know that we have a handful of customers that use it, but it never became a highly used feature,” Larsson said via email. “We have no immediate plans to sunset the feature, but it’s something that we’ll continually monitor based on whether it looks like mask mandates will return or not.”
In a press release last October, Rhombus claimed to count school districts, healthcare providers, city governments, and Fortune 500 companies among its customers.
Alerting features like those that Motorola Solutions and Rhombus offer are of concern to privacy experts, who worry that the technologies will normalize greater levels of surveillance — giving managers ammunition to punish disfavored employees. Amazon notoriously uses algorithms to audit warehouse worker productivity at a granular level, dinging workers for spending too much time away from scanning barcodes or sorting products into bins.
Coincidentally, Amazon a year ago made a big to-do out of a technology called Distance Assistant that the company developed to monitor warehouse workers’ compliance with social distancing rules. Distance Assistant remains available for Amazon warehouse managers to use, but it’s no longer required, spokesperson Barbara Agrait told TC via email, reflecting newer guidance from the U.S. Centers for Disease Control and Prevention.
Other vendors like Montevideo, Uruguay-based Tryolabs have applied these types of technologies to more public venues, including brick-and-mortar stores. An AI consultancy, Tryolabs developed a mask detection product dubbed MaskCam that generates statistics about mask usage in real time. Cofounder and COO Ernesto Rodríguez says that interest has faded compared to the pandemic’s early days, but that MaskCam was at one point set up in Bozeman Montana Airport in Belgrade, Montana to count people walking by and determine the percentage of them wearing a mask.
“Since the interest in this particular solution is declining, Tryolabs continues to focus on researching other problems based on the same core technologies. The models and technologies developed for this particular solution can be applied to other use cases,” Rodríguez told TC via email. “The same libraries can be used and customized in other visual AI solutions, such as in retail, to count the number of people entering and exiting physical stores, or in logistics and supply chain for predictive maintenance scenarios.”
Mission creep has been one of the defining themes of the pandemic where it concerns the tech industry, as evidenced by the sales of facial recognition temperature kiosks with dubious effectiveness (not to mention location-tracking apps). If it wasn’t clear before, hindsight reveals that mask detection was a trojan horse for more problematic technologies, including surveillance technologies, in the workplace and elsewhere.
As the American Civil Liberties Union notes: “Overbroad efforts to curb and track COVID-19 leave the door open to an abiding surveillance apparatus that won’t be dissolved once the public emergency dust settles … We have a duty to ensure that temporary COVID-19 data surveillance infrastructures do not take hold to outlast the effects of this once-in-a-century pandemic.”
The need for face mask detection — if there ever truly was one — will eventually go away. But customers who purchased the technology might be inclined to keep it for less ethical purposes.
