As local and federal authorities across the U.S. face growing restrictions on facial recognition technology, many are turning to a controversial new AI tool that skirts legal limits by tracking people using non-biometric attributes instead. The technology, known as Track, was developed by video analytics firm Veritone and is already in use by more than 400 agencies and institutions—including local police departments, federal prosecutors, and universities.
According to a recent report by MIT Technology Review, which first brought the tool to public attention, Track allows law enforcement to monitor individuals based on observable traits such as body size, clothing, hairstyle, gender, and accessories. The system does not rely on facial features, allowing it to operate in jurisdictions where facial recognition technology is restricted or banned.
Ryan Steelberg, CEO of Veritone, told MIT Technology Review that Track was designed specifically to help authorities “potentially identify criminals” when facial recognition is not permitted or faces are obscured. The software analyzes recorded video and can compile timelines showing a person’s movements across multiple locations and video sources—from body cameras and drones to public footage and citizen uploads from platforms like Ring or YouTube.
Steelberg described the product as the company’s “Jason Bourne app,” capable of constructing complete surveillance narratives without needing a clear view of a subject’s face.
Track’s ability to bypass facial recognition laws by avoiding biometric data has raised significant concerns among civil liberties advocates. While biometric data—such as fingerprints and facial features—is specifically regulated under many state and local laws, attributes like clothing or body shape often fall outside the legal definitions, despite their ability to persistently identify individuals over time.
“This is a potentially authoritarian technology,” said Jay Stanley, a senior policy analyst at the American Civil Liberties Union (ACLU), in an interview with MIT Technology Review. “It gives great powers to the police and the government that will make it easier for them to solve crimes—but also easier to overuse and potentially abuse.”
The ACLU expressed alarm that Track represents the first known use of a non-biometric surveillance system deployed at scale in the United States. The organization also warned that it could normalize mass surveillance under the guise of technological workaround.
While Veritone says only 6% of its business currently comes from the public sector, that segment is its fastest-growing market. Track is already being used in states such as California, Washington, Colorado, New Jersey, and Illinois. At the federal level, the Department of Justice began using the tool in criminal investigations as early as August 2023. Veritone also says its broader suite of AI surveillance tools—including facial recognition—is in use at the Department of Homeland Security and the Department of Defense.
Veritone did not specify which divisions within these federal agencies use Track, and neither the Department of Justice, Homeland Security, nor Defense responded to MIT Technology Review’s requests for comment.
Critics argue that Track significantly expands the scope of video footage that can be used in investigations. Unlike facial recognition, which requires visible, unobstructed faces, Track can analyze video where such detail is unavailable. This means vast archives of video—previously of little forensic value—could now be mined for investigative leads.
“This creates a categorically new scale and nature of privacy invasion,” said Nathan Wessler, an ACLU attorney. “You’re not just speeding up what a cop could already do—you’re creating a capability no human investigator ever had.”
The emergence of tools like Track coincides with the proliferation of laws limiting or banning facial recognition across the U.S. Cities like San Francisco and Oakland have nearly total bans, while states such as Montana and Maine sharply limit its use, especially with live surveillance feeds.
However, many of these laws focus specifically on “biometric” data, leaving room for tools like Track to operate in a legal gray zone. Even so, experts warn that attributes such as body type or recurring clothing items can function as de facto biometric identifiers over time.
Steelberg told MIT Technology Review that Track may soon be able to process live video feeds, a development that would further blur the lines between passive analysis and active surveillance. He also acknowledged that video evidence from Track is already being used in multiple ongoing legal cases, although he declined to provide specifics.
Despite Veritone’s insistence that Track is merely a “culling tool” to help investigators locate relevant footage more efficiently, privacy advocates see a troubling trend.
“Even if it’s not facial recognition, it can achieve many of the same ends—and it’s happening without the same level of oversight,” said Stanley of the ACLU.
As this technology spreads, civil liberties groups are calling for new regulations to ensure that non-biometric AI surveillance tools like Track are subject to the same scrutiny as their biometric counterparts.
