Make cop cameras automated

In the months since the police killings of Eric Garner and Michael Brown, the idea of outfitting police officers with body-mounted cameras has quickly moved from notional to inevitable. Encouraged by a public outcry for more accountability and a $263 million earmark from President Barack Obama, police departments across the U.S. have begun adopting on-body recording systems at a breakneck pace, leaving unanswered questions about what should be recorded and who gets to see the footage.

It’s a conundrum for the digital age: How do you balance the need for police transparency with the privacy rights of Americans who will inevitably be captured by ubiquitous cameras?

On the one hand, robust coverage is needed if there’s any hope the cameras will hold officers accountable. But there’s no quick and easy way to attain that transparency without potentially putting the privacy of victims and innocent bystanders at risk — especially if the police keep massive video archives that can be algorithmically searched and scrutinized later.

As I’ve written before, there are larger issues at play that cop cameras cannot and will not address. They fail to tackle the bigger picture of institutional racism, our broken justice system and the inherent bias that comes from creating an official version of events from the officer’s perspective. But if we assume body cams are inevitable — and that now seems like a fair assumption — the crucial ingredient will be automating as much of the process as possible, removing cops from the equation completely.

Automation must be accomplished in two areas: the act of recording and the storage of files.

In a 2013 paper, the American Civil Liberties Union stressed that officers with body cameras should be recording at all times to prevent selective editing of their encounters. This is hardly universal practice. For example, the Sonoma County Sheriff’s Office in California gives officers broad discretion in when to start and stop recording, according to policy documents obtained through a public records request. The policy doesn’t specify any consequences if deputies fail to record an encounter and even excuses them from doing so whenever they feel it might “jeopardize their safety.”

A report on body cameras (PDF) released last year by the nonprofit Police Executive Research Forum shows that giving officers power over when to record is the norm. Most of the police department body cam policies reviewed “give officers the discretion to not record when doing so would be unsafe, impossible or impractical.” While obviously meant to ensure officer safety, such broad exemptions create an obvious loophole whereby cops can simply state they felt threatened and chose not to record. Further, the report found that of the 63 responsive police departments currently using body cameras, nearly one-third had no written policy governing their use.

Given what we know about the state of law enforcement, the police cannot be trusted as the sole masters of these body cameras. From “testilying,” the widespread practice of police committing perjury without consequence, to the scathing report released on March 4 by the Department of Justice revealing the systemic and destructive persecution of black residents by the Ferguson, Missouri, police, we simply cannot expect cops to be responsible stewards of the cameras or what they record. At the same time, recording everything and keeping it for extended periods widens the dragnet and threatens peoples’ privacy.

This is where automation comes in. If recording all the time is not practical or desirable, all body cameras must be made to automatically start recording whenever a weapon is drawn. In January the Los Angeles Police Department ordered a new line of smart weapons from the stun gun company Taser that does something similar. The camera system has a 30-second rolling buffer: Whenever the weapon is used, the events 30 seconds prior to the moment the camera was triggered are recorded as well. The system keeps a log of the time and duration of every weapon usage.

This functionality must eventually extend to guns and batons so that every police use-of-force event can be accounted for. Crucially, the complete logs must be publicly accessible and not hidden on some corporate server maintained by a stun gun company.

Releasing the video footage is trickier. Peoples’ privacy can be violated if cops record witnesses or victims inside their homes, for example. And unlike photos and paper documents, redacting sensitive information in video is a time-consuming chore.

But the answer is not to deny access to these records wholesale, as several states and cities have proposed. While it’s true that public records laws in some states might make sensitive footage of police interactions available by request, we should probably be less concerned about nosy neighbors and more worried about how the police will use the video they collect behind closed doors.

The Seattle Police Department appears to be finding a middle ground. Last month the department began uploading redacted footage from officers’ body cams to a channel on YouTube. The identities of the videos’ subjects are obscured by thick Gaussian blurs, leaving only their ghostly outlines.

While not perfect, this method has some merit. By uploading fully redacted footage, Seattle police are giving the public access to a kind of placeholder that confirms the video’s existence while preserving the privacy of those being recorded. From there, lawyers, journalists and activists could request the unredacted versions, which should match when overlaid with the YouTube versions.

This is just a stopgap, of course. The software being used for the YouTube videos is the result of a hackathon the Seattle police hosted last December, which brought together several dozen programmers, video technicians and public records advocates to solve the problem of quickly redacting body cam video for release to the public. Eventually, the goal is to automatically redact faces from the footage and leave the rest of the video intact.

There’s just one problem: The Seattle police still have complete control over the source footage. That means there’s no guarantee they aren’t curating the videos or editing them before uploading.

One solution is to have video automatically redacted and uploaded and the unredacted source footage sent securely and automatically to an independent, publicly accountable third party, such as a judge. Police would never touch the footage unless it became relevant to a specific criminal investigation based on reasonable suspicion. This would both prevent the video from being misused for mass-surveillance purposes and make the untainted source material available to police and lawyers on both sides. The independent court could also handle media requests for the footage, testing on a case-by-case basic whether the public interest prevails over individual privacy rights.

Either way, the system won’t be perfect. Like any medium, video is about perspective, and there will always be a subjective difference between putting cameras on cops and having cops on camera. But if ubiquitous police body cameras are inevitable, we must implement technical controls in tandem with policy constraints to ensure that people’s safety and privacy are protected. Without both, cop cameras become just another form of mass surveillance that serves the police more than the public.