A man and a woman were pulled over near 19th Avenue and Broadway Road, in south Phoenix, on 5 August. They say the two officers beat them and used a taser on them. There is no video, because the cameras were never switched on.

Their neighbours are the second set of people in this. Every one of them now has to decide whether calling the police means being believed or being afraid, and they will decide it without knowing whether what happened on that street corner is rare.

Because nobody counts it.

Nothing that was missing caused this

Start with what the officers did not do. They did not activate their body-worn cameras. They did not file a use-of-force report. They did not notify a supervisor by radio.

All three were already required.

"That is not a policy gap; that is an enforcement gap inside an existing policy."MiniMax

That distinction runs through everything below, and it is the reason most of the reforms proposed for cases like this miss. The obvious fix is tamper-proof cameras that switch themselves on and cannot be disabled. And then:

"Body cameras already existed in Phoenix. The officers chose not to activate them. Tamper-proof technology prevents malfunction; it does not prevent deliberate non-compliance by someone who has already decided to conceal. Your solution assumes the problem is technical. The Phoenix case suggests the problem is motivational."MiMo

The counter-argument is decent — automatic activation makes non-activation itself mechanically detectable, so the concealment becomes the evidence. That is real, and it closes something. It also arrives after the stop.

Which sends the question one step earlier, to the part of this nobody usually examines.

It began with a decision nobody had to justify

"The Phoenix case began with a discretionary stop for a minor infraction. The foundational reform isn't about recording that stop perfectly; it's about minimizing such discretionary, officer-initiated encounters."MiMo Flash

Route routine traffic enforcement to civilian agencies. Issue citations from plate readers by post. Take the armed confrontation out of the administration of a minor infraction, and the encounter where this went wrong stops happening.

It is the most structural idea on offer, and it drew the most careful objection — from the side that had spent the whole argument on who gets hurt:

"Minimizing stops through automated citation may remove a badge from the encounter, but it can impose fines, late fees, and license suspensions that fall hardest on low-income and minority drivers. That is a different kind of harm."DeepSeek

Nobody resolved that, and the honest reading is that it is not resolvable in the abstract. It depends on numbers about who gets fined and who loses a licence — which brings us to what this whole argument is actually short of.

The number that does not exist

The figure that gets quoted in every story about police accountability comes from a database kept at Bowling Green State University: 21 officers charged with murder or manslaughter in 2021, the highest since counting began in 2005, against 16 the year before, and 12, 10 and 7 in the three before that. A 200% rise across four years.

It is a real number, carefully kept, and it does not apply here.

"The Phoenix case falls outside this dataset entirely, which means we have zero national-level data on non-lethal on-duty misconduct cases."GLM

That series counts charges arising from on-duty fatal shootings. Nobody died on 19th Avenue. So the most-cited measure of American police accountability is silent on assault, on unreported stops, on cameras that stay off — and no equivalent series exists for any of them.

Nor for the things that would tell you whether Phoenix is an outlier or a pattern: how often body cameras go unactivated when policy requires them, how many of those charges end in conviction, what share of misconduct cases are established without video, as this one was.

So the argument about whether the system is "sufficient" is being had with one measurement that covers a different category of event, and no measurement of the category in front of us. Even the one piece of hard timing available — roughly seventeen days from the stop to the arrest and firing, twenty-one to the charges — is uninterpretable:

"Without comparative data from other departments or prior years, we cannot say whether the current system is 'sufficient' or 'insufficient' — we can only say it produced an outcome in this instance."GLM

What the system did do, and why that is not the argument

It is worth being fair to Phoenix, because the sceptical case was made properly.

Two officers were arrested and fired inside three weeks. The county attorney filed nine felony counts, including kidnapping, aggravated assault, aggravated robbery, tampering with evidence — and conspiracy to hinder prosecution, which is the charge that says this was not one bad minute but an agreement afterwards. The department is auditing the entire precinct. Investigators built probable cause without any video at all.

"The fact that neither had prior disciplinary history is not, on its own, proof that the system is broken. It is proof that pre-employment screening and prior review did not flag these particular individuals — a distinct problem from institutional non-response."MiniMax

That is right, and it is also the whole difficulty. What worked here worked afterwards. Firing, charging and auditing are what a system does once the harm has happened, and by then the man and the woman in that car have already had the thing done to them that the system exists to prevent.

"A quick firing and arrest after the fact is not prevention; it is repair after damage is done. Legal charges do not restore a person's sense of dignity or make a community whole."DeepSeek

What is not known

Conviction rates for charges like these. National figures for non-lethal on-duty misconduct. How often cameras go unactivated against policy. How equitably automated enforcement falls across neighbourhoods. Whether seventeen days is fast.

Five gaps, and they share a shape: each one is the number you would need to know whether to change anything. The single figure that does exist counts only the cases where somebody died — which is the one category serious enough that it was never going to be ignored anyway.

Two officers with clean records skipped three rules that already applied to them, on a stop nobody made them make. The rule was not missing. The count is.