Now, AI Writing Police Reports: What Could Go Wrong?

EFF’s Matthew Guariglia and Dave Maass have brought to light a development in police stations across America that should concern every criminal defense attorney.

A new artificial intelligence program sold by Axon – the manufacturer of police body cameras and tasers – now offers police departments an AI agent that can take audio from body-worn surveillance cameras and convert it into a police report.

Getting police reports right is critical to ensuring justice in the courtroom. These reports are often the first drafts of a courtroom prosecution of a criminal suspect. The officer writes down details about witnessing a possible crime, the discovery of evidence, and reports whether or not the suspect resisted arrest. A prosecutor will scour the police report for details to craft a narrative of guilt before judge and jury. Thus, the accuracy of a police report can mean the difference between freedom or prison for innumerable defendants.

Axon advertises its product, Draft One, as a convenient way to streamline this process. It does the desk work so officers can spend more time on the streets. Sounds good, but what could go wrong?

• Any defense attorney will tell you that in the confusion of an arrest, officers will often order suspects to quit resisting arrest when, in fact, the suspect is trying to be cooperative. The straight conversion of the jumbled audio of an arrest risks all kinds of misrepresentations.

 

• EFF writes: “The public should be skeptical of a language algorithm’s ability to accurately process and distinguish between the wide range of languages, dialects, vernacular, idiom and slang people use.”

 

• EFF: “Police officers who deliberately speak with mistruths or exaggerations to shape the narrative available in body camera footage now have even more of a veneer of plausible deniability with AI-generated police reports. If police were to be caught in a lie concerning what’s in the report, an officer might be able to say that they did not lie: the AI simply mistranslated what was happening in the chaotic video.”

The Draft One process, by design, thwarts any attempt at an audit or to determine whether a statement was written by an officer or written by AI. In filing a Draft One report, an officer does sign an acknowledgement that he or she reviewed the report and edited it in accordance with that officer’s recollection.

However, once the text is copy-and-pasted into the officer’s police report, and the window is closed, the original AI draft disappears. There is nothing to stop an officer from just downloading the AI report and making it official.

This aspect of Draft One is a feature, not a bug.

EFF dug up a video of a discussion about this new product. In it, Axon’s senior principal product manager said: “So we don’t store the original draft and that’s by design and that’s really because the last thing we want to do is create more disclosures headaches for our customers and our attorney’s offices.”

More disclosure headaches? Like what really happened in an arrest?

• EFF sums up this position: “Axon deliberately does not store the original draft written by the Gen AI, because ‘the last thing’ they want is for cops to have to provide that data to anyone (say, a judge, defense attorney or civil liberties non-profit).”

​
We are all for ways to make the difficult job of being a police officer easier. It is not too much to ask, however, that drafts be kept of any AI assistance, and that such assistance be disclosed to defense attorneys. One bill before the California Assembly, SB 524, would be a good start by mandating such transparency in America’s largest state by population.

DONATE & HELP US PROTECT YOUR PRIVACY RIGHTS