The privacy issue is that an agent of the public is intruding into private spaces. The body cam footage is just revealing in a more obvious and exploitable way that the problem already existed. Whatever the body cam sees is already being recorded by the mark I cop eye.
I imagine that the privacy issues would be less severe for the recordings where cops knock, present a warrant, and then enter to execute it, because that is the due process instituted for the public interest breaching into private spaces.
Agents of the public should be doing their duties in public.
As for the automating, I'd use glyphs easily recognized by machine vision to mark different areas in the cop shop. Stencil-paint a glyph on the walls of the bathrooms and locker rooms, and the camera can automatically tag that video as not immediately available for public release, and blur an area surrounding the glyph for automated redaction.
If people are concerned about body cam video of themselves being released automatically without redactions, they can put up glyph stickers in their homes, or wear glyphs on their clothing. A public-private toggle switch on the cam might help, but is susceptible to bad user input or insufficient usage training.
Without some kind of explicit marker, AI may never be able to determine the transition points between public and private spaces, and certainly could not deal with adversarial attacks against that classifier. At worst, over-glyphing or inappropriate wearer inputs could exhaust the video-review resources.
But I think that it is in the public's interest for all calls to the emergency lines to be public, regardless of the potential privacy issues. Instead of having the 911 call center employees worried about covering their asses in case they make a mistake, support them with good training, checklists for the most common issues, a second tier of higher-trained crisis managers to join in for non-routine calls, and maybe an on-demand pool of expert support, such as the psychiatrists, lawyers, or explosives disposal technicians.
Leaving the postmortem analysis of an emergency only to officially authorized people allows cover-ups or deceptive re-framing to happen.
My neighbors left their apartment door open and did not respond to my knocks or my voice. I called the cops, because I wasn’t going to enter the apartment. The cops entered the apartment and found my neighbors in bed. Whether they were asleep or in the throes of passion, I don’t know. But if the cops had had body cams, it would have been captured.
I’ve heard a story about the cops being called by neighbors because someone was screaming for help. Turned out the lady of the house was doing the screaming. She was tied up in bed and the husband was in costume and collapsed on the floor having a heart attack.
These were both legal entries into private places. I think most people would agree these are things the public doesn’t need access to if they are captured by body cams.
Yup. Cops see a lot of things that shouldn't be open to FOIA requests. There's also the little issue that cops are human--they use the bathroom on occasion. If the camera might be running (they're human, they might forget to turn it off) and the video could be FOIAed they're not going to be welcome.
I think a reasonable compromise on this would be to modify the FOIA with regard to things like bodycam video and 911 tapes. Anything which is to be released gets edited for stuff that shouldn't be released and the requestor pays a fixed hourly rate for this.
None of the body cam footage I have seen has ever shown the wearer's own crotch. But this is the perfect use case for manually or automatically tagging inside-bathroom footage as private.
In any case, I don't think that the possibility that someone might skeeze out on public-record video overrides the public interest in having unrestricted access to those records.
In the ancestor post example of the neighbors with the open door discovered in bed, if video existed, it would surely show the cops loudly announcing at the front door before entering the residence, and continually announcing their presence while inside, before encountering the occupants. That's an attempt to salvage privacy, as well as a reasonable thing to do to determine if anything is wrong inside the residence.
I think a reasonable expectation of privacy might necessarily include closing your front door. And an open front door in a dense residential area is reasonable grounds to conduct a warrantless search. Urban residents usually don't want strangers or neighbors just walking right in to their homes without an explicit invitation, so doors stay closed and locked. In some states, entering an open residence uninvited [during daylight hours] is not considered trespassing. That requires an explicit act by the residents to inform guests that they are not welcome inside, such as closing and locking the front door, or telling them to leave.
It's the reasonable expectation of privacy that's the key. And even bench judges sometimes struggle with it. There's no chance that an AI could make that determination at the current level of development, including as it does factors such as time of day, character of the neighborhood, and local law.
It could detect and redact certain body parts, especially faces. The auto-redacted videos could then be released immediately. If an unredacted version of a video is requested, the video review officer could independently remove each redaction box using custom software tools, based on a checklist or flowchart of guidelines. I'm not quite certain as to whether the requester should be required to pay the full costs of un-redacting the video or not. Certainly, only the first requester should pay it, if anyone must.
I imagine that the privacy issues would be less severe for the recordings where cops knock, present a warrant, and then enter to execute it, because that is the due process instituted for the public interest breaching into private spaces.
Agents of the public should be doing their duties in public.
As for the automating, I'd use glyphs easily recognized by machine vision to mark different areas in the cop shop. Stencil-paint a glyph on the walls of the bathrooms and locker rooms, and the camera can automatically tag that video as not immediately available for public release, and blur an area surrounding the glyph for automated redaction.
If people are concerned about body cam video of themselves being released automatically without redactions, they can put up glyph stickers in their homes, or wear glyphs on their clothing. A public-private toggle switch on the cam might help, but is susceptible to bad user input or insufficient usage training.
Without some kind of explicit marker, AI may never be able to determine the transition points between public and private spaces, and certainly could not deal with adversarial attacks against that classifier. At worst, over-glyphing or inappropriate wearer inputs could exhaust the video-review resources.
But I think that it is in the public's interest for all calls to the emergency lines to be public, regardless of the potential privacy issues. Instead of having the 911 call center employees worried about covering their asses in case they make a mistake, support them with good training, checklists for the most common issues, a second tier of higher-trained crisis managers to join in for non-routine calls, and maybe an on-demand pool of expert support, such as the psychiatrists, lawyers, or explosives disposal technicians.
Leaving the postmortem analysis of an emergency only to officially authorized people allows cover-ups or deceptive re-framing to happen.