The New AI Tools Helping Police Bypass Facial Recognition Bans
By Better Together Agency
Rules written around one technology can leave room for another system to create many of the same privacy and equity concerns. New tracking tools in policing show why AI governance has to focus on capability and impact, not a product label.
Tracking people without facial recognition
MIT Technology Review reported on Veritone Track, a video analytics tool that can follow a person across footage using characteristics such as body size, hair, clothing and accessories rather than identifying a face. The company said the tool was being used by hundreds of customers, including law enforcement and universities.
The distinction matters legally. A rule that prohibits facial recognition may not cover a system built around other physical attributes even when the practical result is persistent tracking.
The governance gap is bigger than one technology
This is a familiar policy problem. Regulation often names the technology people already understand. New systems then arrive with a different technical method but a similar effect on privacy, civil liberties or access.
Responsible AI policy should ask what a system can do, who it affects and what safeguards apply. A definition based entirely on whether software identifies a face can miss tools that identify or track a person by other characteristics.
Bias does not disappear because the biometric changes
Systems that classify people by gender, clothing, body characteristics or other observable traits still depend on categories created by people. Those categories can be incomplete, inaccurate or culturally narrow.
When law enforcement uses those classifications, the consequences of error can be far more serious than a bad recommendation or an awkward chatbot answer. That raises the standard for testing, transparency and human review.
Communications belongs in the governance plan
Public agencies need more than an internal legal interpretation. Communities should be able to understand what technology is being used, what information it analyzes, how long data is retained and what options exist when the system produces a harmful result.
If an agency cannot explain the policy in plain language, public trust will depend on assumptions instead of informed consent and oversight.
Questions policymakers and public agencies should ask
- What capability is the policy trying to regulate?
- Can the system track or classify people without naming them?
- What error rates and demographic differences have been tested?
- What data is retained and who can access it?
- What human review is required before action is taken?
- How will the public be told what the system does?