An artificial intelligence review of Knoxville Police Department body-camera footage has given most officer-public encounters a thumbs up after sifting through nearly 7,000 interactions and more than 2,100 hours of video from last year (2025). The upbeat finding, flagged in a recent local TV segment, is already prompting a more skeptical follow-up question in Knoxville: what can an algorithm really tell the public about policing without showing its work?
The AI assessment, which coded the majority of those encounters as positive, was first reported by WVLT. The short broadcast did not link to a public report or spell out the technical details, so viewers were left without answers on which software or vendor was used, what counted as a “positive” interaction, or how the scores were checked for accuracy.
The Knoxville Police Department already does its own human-powered reviews of body-worn camera footage. In its 2024 “Year in Review,” the agency highlights the creation of a Police Audits and Compliance Specialist within the Office of Professional Standards to examine videos and flag potential training or policy issues, according to the Knoxville Police Department. The department says that role is meant to tighten quality control and address concerns through training, policy changes or other resources, not just punishment.
How automated analysis works
On the federal research front, multimodal AI tools are being designed to watch body-camera footage in much the same way a social scientist would, only at high speed. According to the Office of Justice Programs, these systems combine audio transcription, natural language processing and computer vision to rate officer encounters on procedural justice markers such as respect, explanation and neutrality.
One tool, TrustStat, was built with funding from the National Institute of Justice. In evaluations cited by the Office of Justice Programs, its automated ratings largely lined up with assessments from human coders. At the same time, the researchers stressed that tools like this need to be validated in different cities and departments, and that methods must be transparent if communities are going to trust the scores.
Limits and privacy concerns
National reporting has also thrown some cold water on the AI body-camera boom. ProPublica has documented how departments are leaning on AI to comb through millions of hours of footage, warning that errors, bias and a quiet expansion of surveillance powers are real risks if the technology is rolled out without guardrails.
The Associated Press has reported on experimental projects, including an Edmonton trial in which cameras were trained to spot thousands of people on a watchlist, that sparked intense debates over ethics and privacy. Those kinds of examples are now part of the backdrop whenever a police department announces new AI tools, even ones framed as internal accountability measures.
In Knoxville, the latest assessment could eventually shape how the department tracks officer performance and designs training. At the same time, it has fueled calls for the underlying data and methodology to be released so outside researchers and community members can vet the findings for themselves. Until that happens, experts say the upbeat numbers should be treated as an opening move in a longer public conversation, not proof that the culture of policing has definitively changed.
