A new form of AI video analysis is giving police and federal agencies another way to follow people through footage, even when facial recognition is restricted or faces are not visible. The tool, called Track, was built by Veritone and can identify people through details such as body size, gender, hair color and style, clothing, shoes, and accessories.
That capability matters because laws limiting facial recognition have spread in parts of the US. Track does not rely on a clear view of a face, which makes it useful for investigations but also makes it controversial. Civil liberties advocates warn that it could create a large-scale tracking system that avoids some rules written for facial recognition.
What Track Does
Track is part of Veritone’s video analytics work. The company says it is used by 400 customers, including state and local police departments and universities across the US. US attorneys at the Department of Justice began using Track for criminal investigations last August.
Veritone’s broader AI suite, which includes facial recognition, is also used by the Department of Homeland Security and the Department of Defense, according to the company. The Department of Homeland Security includes immigration agencies.
Track works by analyzing recorded video and comparing visible traits. A user can search footage for a person based on combinations of attributes, then use the tool to assemble timelines across different locations and video feeds. Veritone demonstrated the system on footage from environments including the January 6 riots and subway stations.
The tool can be accessed through Amazon and Microsoft cloud platforms. Agencies can add video from several sources, including police body cameras, drones, public videos on YouTube, and footage uploaded by citizens from Ring cameras or cell phones in response to police requests.
Why It Changes The Facial Recognition Debate
Facial recognition generally depends on footage where a face is visible and usable. Track does not have that same limitation. That difference could expand the amount of video available to investigators, because a person can potentially be followed through clothing, body size, hair, shoes, or accessories.
Ryan Steelberg, Veritone’s CEO, described the original idea behind Track as a way to help when agencies are not allowed to track faces. He also said the tool can help when faces are obscured or missing from the footage.
Veritone says Track currently runs only on recorded video. Steelberg claims the company is less than a year from being able to run it on live video feeds. He also said the number of attributes used by Track will continue to grow.
A company spokesperson said skin tone is one of the attributes the algorithm uses to tell people apart, but that the software does not currently let users search by skin color.
Why Civil Liberties Groups Are Alarmed
The American Civil Liberties Union raised concerns after learning about Track through MIT Technology Review. The group said it was the first instance it had seen of a nonbiometric tracking system being used at scale in the US.
Jay Stanley, a senior policy analyst at the ACLU, had written in 2019 that artificial intelligence could one day make it much easier to search surveillance footage automatically, even when no crime had happened. He now says Track is the first product he has seen that makes broad tracking of particular people technologically feasible at scale.
Stanley called the technology potentially authoritarian and warned that it could help police solve some crimes while also making overuse and abuse easier. The concern is not only that the software speeds up existing work. The concern is that it may make a new kind of broad surveillance practical.
Nathan Wessler, an attorney for the ACLU, argues that Track can create a scale of privacy invasion that was not previously available to police. If officers can search footage they would not have reviewed before, the practical reach of surveillance grows.
The Legal Gray Area
Track is expanding while facial recognition limits are spreading. The source article notes that laws in Montana and Maine sharply limit when police can use facial recognition and do not allow it in real time with live video. San Francisco and Oakland, California have near-complete bans on facial recognition.
Those rules often refer to biometric data, but Wessler says the phrase is not clearly defined. It generally points to characteristics that do not change, such as faces, gait, and fingerprints. Clothing changes, but some traits and habits blur the line.
For example, Wessler points to a person in winter who often wears the same boots, coat, and backpack. If that person’s visible profile remains consistent across many saved video feeds, the practical effect may look similar to face recognition, even if the system is not technically identifying a face.
That distinction is central to the debate. Track may raise many of the same privacy questions as facial recognition while falling outside some rules that were written specifically for facial recognition or biometric data.
Where The Technology May Go Next
Veritone says the public sector is currently only 6% of its business, with most clients in media and entertainment. Even so, the company says the public sector is its fastest-growing market, with clients in places including California, Washington, Colorado, New Jersey, and Illinois.
Jon Gacek, general manager of Veritone’s public-sector business, said Track is a culling tool meant to speed up the task of finding important parts of videos, not a general surveillance tool. Veritone did not specify which groups within the Department of Homeland Security or other federal agencies use Track.
The Departments of Defense, Justice, and Homeland Security did not respond to requests for comment. Steelberg said several ongoing cases include video evidence from Track, but he could not name the cases or comment further.
For now, one important question remains unresolved: whether Track is being adopted in places where facial recognition is banned. The answer matters because the technology could become a workaround for restrictions that were meant to limit government tracking of individuals.