Soon, thousand of police officers across the country will don body-worn cameras when they go out among the public. Those cameras will generate millions of hours of footage—intimate views of commuters receiving speeding tickets, teens getting arrested for marijuana possession, and assault victims at some of the worst moments of their lives.

As the Washington Post and the Associated Press have reported, lawmakers in at least 15 states have proposed exempting body-cam footage from local open records laws. But the flurry of lawmaking speaks to a larger crisis: Once those millions of hours of footage have been captured, no one is sure what to do with them.

I talked to several representatives from privacy, civil rights, and progressive advocacy groups working on body cameras. Even among these often allied groups, there’s little consensus about the kind of policies that should exist around releasing footage.

Body cameras were introduced as a tool of public accountability, but making their videos available to the public might be too fraught, too complex, and too expensive to actually put into practice.

* * *

Much of the ambiguity around body cameras comes down to this: Despite their general popularity, despite being the only policy change called for by the family of Michael Brown, body cameras are a little weird. They are both a way for the public to see what police officers are doing and a way for people to be surveilled. If a body-cam program, scaled across an entire department, were to release its footage willy-nilly, it would be a privacy catastrophe for untold people. Police-worn cameras don't just capture footage from city streets or other public places. Officers enter people's homes, often when those people are at their most vulnerable.

So while body-cam footage is “very clearly a public interest record,” says Emily Shaw, the national policy manager at the Sunlight Foundation, it is also “just full of private information.”

What’s more, there’s no easy way to fix this. Normally, private information would be redacted out of public records. (The Sunlight Foundation recommends certain easy ways to do this for other kinds of records, such as putting all personal information-containing columns off to one side of a spreadsheet.) But there’s no simple way to compartmentalize the sensitive aspects of video footage. Video isn’t text: It’s dozens of frames per second, each its own potentially private visual. Redacting personal information out of the frame remains both expensive and unreliable, and it might not be possible with algorithms.

“If you just put a swirl blur on somebody’s face, it’s not very difficult to unswirl that blur, and then all of a sudden it’s un-redacted,” Shaw said.

The only effective redaction process might be for technicians to go through a video frame by frame—an exceptionally costly process. Even that might not be enough. “If you have additional data, then you can re-identify people who’ve been anonymized or intentionally anonymized,” Shaw said.

That technique is so prohibitive that some cities have declared the whole situation impossible and blanket-exempted footage from open-records requests. Others are attempting technical solutions: Seattle’s police department is trying to “auto-redact” videos, muting a video’s sound, blurring the picture, and posting it to a YouTube channel.

But some experts say that, if departments can’t deal with the high cost of Freedom of Information Act (FOIA) requests, then officers shouldn’t get body cameras.

“If the police cannot handle FOIA requests for body-cam footage, they should not use them,” said Jeramie Scott, a national security counsel at the Electronic Privacy Information Center (EPIC). “There should be no blanket FOIA exemptions for police body cameras.”

Scott thinks existing exemptions, which cover privacy and reducing potential harm to private citizens, work just fine for body cameras.

* * *

If there’s any consensus among experts considering body camera policy, it’s about this: Most groups, including the ACLU, agree that individuals recorded by body cameras should have access to that footage.

Yet implementing even that provision is tricky, says Shaw.

“If there are several people in the video and some of them don’t want it to be public but one of them really does, what happens in that circumstance?,” she told me. “Even if you redact the person who doesn’t want to be seen there, everybody knows this person is an associate of this person who is visible.”

Alex Rosenblat, a researcher at Data and Society, said a policy could have particularly “shady consequences for the victims of domestic violence or sexual assault.”

“What happens when the perpetrator (who may not be prosecuted, particularly if that's not in the best interests of the involved parties) is an additional subject of footage that involves a victim?” said Rosenblat in an email.

Rosenblat added that most conversations about police body cameras concluded that subjects of police recordings who would later regret such a recording need to ask officers to turn their cameras off. But, she said, “that puts the onus of anticipating the downstream effects of police-collected recordings and access issues onto an individual who might not be savvy to them.”

What’s more, police officers themselves are often the subjects of videos too. If a law gives all subjects of recordings unfettered access to them—like Michigan’s proposed body-cam law—does that mean that officers have access to the videos, too? “This would run contrary to the positions of many civil rights advocates, who would support policies that restrict who has access to video internally,” said Rosenblat. The Leadership Conference on Civil and Human Rights, for example, believes that officers should not have access to footage before they write an incident report.

And while most experts seemed to think that some if not all of the questions about granting subject access to footage will be resolved, it only stands to get more complicated. If police work at a school or hospital, how should that footage be considered? “In the case of schools, it would probably hinge on privacy protections offered to minors, and on whether police body-camera footage in schools are considered student data,” said Rosenblat.

* * *

In this case, laws related to the mass-implementation of body cameras may be outpacing the experts. In the District of Columbia, Mayor Muriel Bowser has moved to exempt all body-camera footage from the city’s open-records law, weeks after promising a new era of accountability and transparency.

Other city leaders have disagreed.

“They are the people’s records. The people bought them,” Delroy Burton, chairman of the city’s police union, told me. Burton said he believes existing open-records limitations are appropriate for body-camera footage. He blamed the mayor’s exemption proposal on her acceleration of the city’s pilot body-camera program.

“I think she was trying to make brownie points, or to score political points—to take what was a three-year plan initially when it was announced” and speed it up, said Burton. (Bowser’s office did not immediately reply to a request for comment on Friday.)

But that may be exactly the problem. It is politically trendy to support body cameras. They have triumphed because cameras sound great in theory. Supported both by officers and reformists, they promise to keep both citizens and cops accountable. But they’re no longer theoretical: Cameras are working tools, riddled with complexities, subject to power. Everyone might support body cameras in theory. No one might support them when they need to implemented in the real world of technology.