It’s budget season, and an artificial intelligence (AI) application for writing up police reports from body-worn camera and drone data is available from one of DeKalb Police Department’s favorite vendors. Does it lurk in the PD’s budget worksheets? If so, the city should start public conversations now about oversight, transparency, and other pros and cons of the burgeoning tech.
What’s the Big Deal
Even if you’re not particularly concerned about the surveillance state, it’s reasonable to exercise caution in using new tech in place of human discipline and judgment, especially when the result might show up as evidence in legal proceedings. AI is terrifically error-prone and requires strict limits and oversight. As both the buyers and the targets of city cameras, the general public should have input into their use. But as we’ll see a few paragraphs from here, DeKalb has already managed to botch it at least once.
The Vendor & the Product
It’s Axon Enterprise, formerly Taser International. DeKalb still buys Tasers and Taser accessories from Axon, and in 2020 deepened the relationship by rejecting a lower bid and approving a $415,000, 60-month contract to kick off its body-worn camera (BWC) program.
DeKalb’s BWC package includes Evidence.com, a cloud-based digital records management platform. The new AI, introduced last year, is an add-on to the Evidence platform and called Draft One. Draft One can automatically generate incident reports based on data from Axon’s BWC and drones — images, camera footage, audio — that are stored in Evidence.com.
The city’s response to a request under the Freedom of Information Act shows Axon has been steadily advertising Draft One to DeKalb Police this year via email, and a couple of officers have signed up for a webinar about this product.
It seems likely that DeKalb will contemplate the purchase of Draft One or a similar AI product at some point. Smart policy development for its use should precede purchase. But the city no longer always solicits public input as it ought.
I am Disappoint
DeKalb’s plan in 2020 to buy body-worn cameras (BWC) didn’t generate much noise, at least not on the record. But approve or oppose, there was a record. Everyone had an opportunity to speak publicly on it, because the city had placed the item on a city council agenda.
The city went even further — laudably — to gather public input for the intended purchase of automated license plate reader cameras. Before the council discussed and voted, DeKalb’s Human Relations Commission (HRC) was tasked with vetting proposed guidelines for their use.
However, neither HRC nor council held public discussions about the city’s drone program. DeKalb put drones into place in 2023 with zero public examination of the implications for privacy and other rights — a significant and unacceptable break with precedent when it comes to introducing surveillance tech into our community.
The reason for the failure to vet the drone program is simple: the drones didn’t cost enough to trigger the requirement. Total costs under the current 5-year drone contract are a bit more than $25,000, but yearly expenditures come to less than $10,000; only if a budgeted purchase reaches a $20,000 threshold for the current fiscal year does it require separate and explicit council approval.
The lapse signals that spending authority isn’t enough anymore. Council should create an additional trigger for requiring public airings of proposals that potentially affect privacy, due process, etc., regardless of cost.
Initiating the AI conversation
Assuming DeKalb PD is or will be interested in Draft One or other AI, the question now is whether planned usage gets publicized and examined automatically, or if we need to apply pressure.
With the city’s regular Committee of the Whole meetings discontinued, and the current mayor’s continuing failure to appoint members to the Human Relations Commission, the number of opportunities for dialogue is greatly reduced.
As things stand now, it would come down to pricing and/or how much equipment is involved, as it did with the drone program. DeKalb’s body-worn camera (BWC) contract expires near the end of this year. If BWC upgrades are contemplated, it’s possible we’ll see a proposal to bundle the Draft One app into a new multiyear contract, in which case the cost would be likely to meet the threshold for requiring council involvement. On the other hand, DeKalb already has Axon BWC, an Axon drone, and licenses for the Evidence.com platform. With all that hardware and software on hand already, a purchase of the Draft One add-on (plus a transcription program if needed) might avoid the trigger for its appearance in an open meeting.
But maybe we shouldn’t wait to see what happens. If it’s not Axon’s AI, it’ll be Flock’s at some point. Here are some positions and steps you can take to encourage greater transparency and opportunities for public input into tech usage:
— Support the return of Human Relations Commission and of council’s regular Committee of the Whole meetings for more thorough discussion of these topics without the pressure of immediate votes.
— Support the development of criteria that would trigger council/committee discussion on tech usage with potential privacy and other rights issues, regardless of the cost of the tech.
— Share your positions on social media, at city council or finance committee meeting, and/or with your council members via phone or email.
I will discuss AI guardrails in a separate post, probably at the City Barbs Blog Facebook group.
Related
Electronic Frontier Foundation: What can go wrong when police use AI to write reports?
The Yale Herald: A digitized New Haven Police: Investigating Axon’s Draft One
ACLU: White paper on police departments’ use of AI to draft police reports