On Wednesday, Axon (formerly “Taser”) announced its offer to outfit every cop in the US with a free body camera, with rollout beginning as soon as the end of the month. About 20 per cent of US police departments use body cameras. The overwhelmingly majority of all police departments have no policies about how best to use the cameras, what to do with footage, or even when to record.
Privacy experts are concerned that embracing this technology without regulation only undermines its original goals of transparency and accountability. What’s more, unregulated introduction of technology into the police force would result in a variety of unprecedented legal and safety issues.
Harlan Yu is a principal researcher for Upturn, a tech policy nonprofit that produced the Body Worn Camera scorecard — a 2016 report rating body camera policies for different police departments. Upturn found that for the few existing body camera policies in the United States, there is no consistent standard. Additionally, the majority of police departments only vaguely address crucial aspects of body camera use, such as personal privacy concerns, and some — such as unnecessary footage retention — are barely addressed at all.
“Axon’s offer creates a perverse incentive for departments to rush into deploying body worn cameras without taking the necessary time to engage with the community and think through many of these hard policy trade-offs before making the snap judgement to go with this free offer,” Yu told Gizmodo. When constructing the necessary regulations, the police departments need to be aware that for every new policy, there is a benefit and a loss.
When to record
The first policy hurdle is deciding simply when the cameras record. Yu is opposed to continuous recording, in which officers simply keep their cameras on throughout every shift, because it picks up so much mundane, public data — people walking on footpaths and sitting in traffic, for example. Departments should instruct officers to record the overwhelming majority of public interactions, Yu proposed. He emphasised that the police should also clearly communicate to civilians that they’re being recorded. But, even that has caveats.
When not to record
“I think there should be particular sensitivity for victims to be able to opt out,” Yu said, referring to victims of domestic violence and sexual assault.
Jay Stanley is a senior policy analyst with the ACLU and the co-author of the organisation’s 2015 policy recommendations for police departments utilising body cams. Stanley agrees that in cases where officers respond to domestic violence and sexual assault, particularly when children are involved, officers should defer to victims for consent to be filmed. Similarly, informants and witnesses, who can endanger themselves by helping police, should have similar privileges. The US federal grant program for body cameras encourages police departments to contact victims rights’ groups when creating body camera policies, though Axon’s offer has no such requisite.
“Officers ought to not have very much discretion over when to or not to record,” Yu said. Officers shouldn’t individually decide when cameras their cameras should be on or off — there should be policy for every scenario.
When to record is one problem, but what happens when officers don’t record when they should? In practice, officers have time and again reported that the cameras have either “fallen off” or spontaneously stopped recording before a fatal incident. How should these officers be held responsible? How should they be punished? Most departments don’t have clear-cut guidelines.
“Different departments will have different ways of going about it, but it should be an escalating process,” Yu offered. The ACLU suggests specific guidelines, which also encourage a punishment scale that gets worse for the officer depending on severity of the incident or whether it is a repeated offence.
Reviewing footage before submitting police reports
Some policies require officers to submit their reports before seeing footage. Some actually encourage officers to view footage before completing their report.
Yu suggests a two-step process. First, the civilian and officer involved both provide a written statement immediately following an incident, before seeing the footage. Second, both the civilian and the officer view the footage of the incident and provide written statements addressing the discrepancies between the original statement and what is depicted in the footage.
This process would partially level the playing field. It would also, for example, prevent officers from tailoring their reports to the footage so that any inconsistencies in their statements are not exposed.
Yu discussed concerns of a “chilling effect” on civilians, where they may feel uncomfortable reporting officer abuse because of the veneer of objectivity the legal system gives video footage. If a policy is firmly in place allowing the civilian to give a statement, regardless of the footage, these effects could be partially alleviated.
Retaining the footage
How long should police departments keep footage they aren’t using? If, for example, 100 patrolling officers record an hour of footage daily, that’s 100 hours of footage to review. It isn’t always easy to say “keep what’s important”, because it takes so long to review all that footage.
Departments should be clear in how long data can be kept before deletion. Yu recommends six months, as it “limits the privacy risks of having all that footage around”.
Mining the footage with AI
Placing firm time limits on footage retention “would also limit the ability of departments to mine that footage, especially when were seeing AI technology coming down the pike,” Yu added.
In February, Axon announced it had acquired a computer vision startup, Dextro Inc, which allowed for AI-powered object detection and unprecedented video data search capabilities. There were a number of privacy concerns, such as the technology’s capability for automated public surveillance and biometric tracking. Yu says police departments should make immediate headway in regulating AI and similar metadata analytics with such dangerous potential.
“[It’s] not what communities probably were expecting when they said, ‘OK, let’s adopt body cams’,” Yu said of AI-enabling features like face and object recognition. “This is a feature that is going to get snuck in with body cam technology that I think is a very dangerous combination.”
Stanley says that any type of analytics should only be used for footage flagged for use of force or a specific complaint.
Too much, too soon
Potentially as many as 14,000 US police departments are being offered this technology without any regulatory framework in place and no training beyond that which is offered by the private company.
Police body cameras were originally presented as a solution for persistent and urgent issues, such as lack of transparency and accountability involving incidents of police brutality and fatal shootings, and the subsequent community mistrust of the police force. Yu and Stanley both underscore that body cameras are not a replacement for substantive police reform.
Moreover, adding these cameras before the difficult work of deciding how to balance privacy, safety and security can do more harm than good.