AI is now auditing police bodycams in seconds

Millions of hours of police bodycam video sit in the cloud untouched, only reviewed when someone dies or a formal complaint blows up on the evening news.

Chicago startup Truleo wants to make sure those tapes actually get watched, using AI to audit patrol audio automatically.

Listening in on the beat

Instead of crunching massive video files, Truleo ingests daily audio from body cameras, transcribing and analyzing interactions in seconds using natural language processing.

Developed on a dataset of over 50,000 positive police encounters, the model flags red-flag phrases like "I can't breathe" or "you're hurting me," along with slurs, threats, and physical altercations.

It also monitors command styles. In high-stress stops, officers who give context rather than bark conflicting orders get praised, while excessive shouting triggers alerts.

In the death of Tyre Nichols in Memphis, officers issued 71 conflicting commands in just 13 minutes. Truleo aims to catch that breakdown in communication before it turns deadly.

When an officer stays calm during a tense stop, both the officer and their sergeant receive an email commendation. If things turn abusive, command staff gets notified within hours.

Results are trickling in. In Alameda, California, police saw use-of-force incidents plunge 36%, officer profanity drop 30%, and civilian non-compliance decline 12%.

At roughly $50 per officer per month, the platform has won a two-year renewal from Seattle and adoption across roughly a dozen California departments, with Aurora, Colorado coming on board next.

Why it matters

Truleo founder Anthony Tassone previously ran AI analyzing customer service calls for Wall Street banks. Repurposing call-center speech analytics for public safety is a remarkably clever wedge: it bypasses the heavy compute of computer vision and treats officer communication like frontline operational data.

For supervisors, it replaces random audit lotteries with automated early interventions, and the price tag can easily be covered by municipal insurance discounts or federal community grants.

Still, the ceiling on this tech is defined entirely by human compliance. An algorithm cannot analyze footage an officer forgot or refused to record.

More importantly, police accountability requires leaders willing to act on bad metrics. As Tassone noted, the hardest barrier isn't the AI, but department chiefs who simply do not want to see what is on the tape.