If you’ve ever worried that advanced AI tech could be used to track people more invasively than ever, your fears are partly becoming a reality. Veritone has developed an AI model called Track AI that can identify and follow people in video footage of any kind.
You might think tech like this can’t be legal, since facial recognition for surveillance is banned in many parts of the US. You’d be right, but Track AI doesn’t rely on facial recognition. Instead, it analyzes body size, gender, hair color, clothing, and even how a person walks to tell individuals apart.
Tracking gait and other physical attributes while skipping facial recognition might make Track AI appealing to many law enforcement agencies. Apparently, some are already using the early version of this surveillance tech in their investigations. And yes, Track AI raises privacy concerns, just like facial recognition used for mass surveillance.
That said, the Track AI software could still be a helpful tool for law enforcement, as long as it’s used lawfully. The real worry comes from the human factor. It’s people who could abuse this powerful AI for personal reasons. And let’s not forget authoritarian regimes that could develop similar tech or adopt Track AI to monitor people almost in real time.
A report from MIT Technology Review says Veritone’s AI tech already has 400 customers. That includes several public sector entities in the US, such as state and local police departments, universities, the Department of Justice, the Department of Homeland Security, and the Department of Defense.
Veritone hasn’t disclosed which groups within those agencies are using Track or what they’re using it for. The departments didn’t comment for MIT Technology Review’s report. But Veritone did demo the technology, and the article includes a short video clip showing Track AI in action.
Imagine you’re trying to find a suspect in a batch of surveillance footage. Facial recognition is off the table because of local bans. With Track, you can select various attributes about the suspect. In the demo MIT shared, the investigator chooses an accessory (a backpack) and describes clothing details (gray color and long sleeves for the upper body).
That’s enough for the AI to generate a list of potential matches. The investigator can then review the footage and confirm the correct ones. While Track AI doesn’t use facial recognition, the human reviewing the footage could still recognize the suspect’s face or other features.
Once a match is confirmed, the AI creates a timeline of the suspect’s movements. Just like that, Track AI pulls together a wealth of video data that might help move an investigation forward. Instead of having humans spend hours watching surveillance videos, the AI does the grunt work, and the human just verifies the results.
Privacy concerns aside, this kind of tech could really help in everyday scenarios. We’ve talked about those iPhone thefts where criminals trick owners into entering passwords, then steal their data and lock them out of iCloud. Tech like Track AI could help solve crimes like that—and many others caught on video.
Track AI currently works with any recorded video, whether from surveillance cameras or other sources. But Veritone CEO Ryan Steelberg told MIT that the company expects it to work with live video surveillance within about a year.
Even so, we can’t ignore the privacy risks, no matter how useful the tech may be. There’s a real concern that those with access to Track AI could misuse it. You can imagine how easy it would be for someone with access to surveillance footage to track someone for personal reasons.
There’s also a broader concern that governments might use Track AI for mass surveillance. The ACLU is definitely watching, warning MIT that this is “potentially authoritarian technology” coming from Veritone.
Jay Stanley, a senior policy analyst at the ACLU, told MIT that Track AI gives police and government agencies powerful new tools. While it might help solve crimes, it also makes it easier to overuse or abuse the technology.
ACLU attorney Nathan Wessler pointed out that Track AI doesn’t have the same limitations as facial recognition. The latter needs clear shots of faces, but Track AI can identify people by how they walk. That means law enforcement can now make use of video footage that might have been useless before.
The report also says that most of Veritone’s customers are in the private sector. Only 6% are from the public sector, but that segment is currently the company’s fastest-growing market.
As mentioned earlier, it’s not surprising to see AI used for surveillance, given recent developments from AI firms. Tools like ChatGPT and Gemini can now recognize objects and locations in images, so it was only a matter of time before companies started building AI tools for law enforcement, and only a matter of time before we started worrying about privacy.
Track AI likely won’t be the last tool to raise these concerns. The good news is that democratic governments will probably put privacy protections in place to ensure the tech is used to catch criminals, not for widespread surveillance.
In the meantime, we need to stay informed about what tools like Track AI can do. Be sure to check out the MIT Technology Review report and watch the video demo to see just how easy it is to use this tech to track people in videos.
