About the data
This site republishes drone flight logs that police and public-safety agencies have already made public. It does not change them, and it does not obtain anything that was not already published.
Where it comes from
There is no national register of police drone flights, and no law requiring one. What exists is a scattering of dashboards, each run by a different vendor, each publishing a different subset of a different set of fields. Finding them is most of the work. This site reads five kinds of source, and treats each one’s limits as part of the record rather than smoothing them away.
Skydio transparency dashboards — 102 agencies, 102,052 flights. Agencies that fly Skydio drones can publish selected flights to a public dashboard. Skydio hosts every one of those dashboards as a public map service, and this site reads those services directly. Only flights on Skydio aircraft appear there, so an agency that also flies other manufacturers’ drones will look smaller here than it actually is.
Flock Aerodome community dashboards — 68 agencies, 99,492 flights. These carry the most detail of any source: the type of call, the agency’s own case number, the priority assigned to it, and a block-level address. Each dashboard shows about a month at a time by default, and there is no index of them anywhere — the ones here were found by searching public certificate records for the hostnames the vendor issues.
Motorola CAPE transparency portals — 5 agencies, 8,746 flights. Each agency chooses how much history to expose. Three publish everything; the rest publish only the last thirty or sixty days, after which a flight disappears from the source permanently. For those agencies this site holds flights the original portal can no longer show.
AirData public flight logs — 20 agencies, 77,211 flights. The oldest records collected here, reaching back to 2019 for Sacramento and 2021 for Chula Vista. AirData publishes no flight duration at all, so those agencies have flight counts but no hours.
San Francisco Police Department — 1 agency, 7,481 flights. SFPD does not publish through a vendor. Under the city’s surveillance-technology ordinance it publishes its own flight log to San Francisco’s open-data portal, and that log covers its whole fleet regardless of manufacturer. It records a date but no time of day, so its page has no hour-of-day view. That is a property of the source, not a defect in the records.
Counting
Every figure here counts distinct flights, never rows of data. That distinction matters more than it sounds: several of the published map services contain duplicate rows — the same flight, with the same takeoff time, landing time and stated purpose, published two or three times over. Cincinnati’s service holds 30,071 rows for 17,938 real flights. Counting rows would overstate the total on this site by roughly fifteen percent.
13 agencies publish the same programme in more than one place. The platforms do not agree about what a flight is called, so a shared flight is recognised by what it describes: the same case number on the same day, or takeoff times within five minutes of each other. Where two platforms describe one flight, the fuller account is kept and the flight is counted once. 200 flights across the site were published twice. Each agency page lists the platforms it draws on.
What is not here
- Agencies decide which flights to publish. Every count is a floor, not a total.
- No flight paths. Several sources publish the full track of every flight as a list of coordinates. None of it is stored here; it is discarded at the moment the page is read, before anything is written down. Each agency page links to that agency’s own official map, which is the authoritative view of where its drones flew.
- Where an agency itself publishes a place for a flight, that text is kept as the agency wrote it — usually a block-level address or an intersection, which is what Flock dashboards show, and never a precise coordinate.
- No pilot names, aircraft serial numbers or dock identifiers, even where a source technically exposes those fields.
- Stated purposes appear exactly as the agency recorded them. Agencies use different vocabularies — some write a statute citation, some an incident type, some nothing at all — so purposes are not comparable between agencies and this site does not try to make them so.
- No national totals or national trends. Aggregate counts across agencies mostly reflect when each agency started publishing rather than how much anyone flies, so a rising national line would say more about dashboard adoption than about drones.
- 44 agencies are collected but not shown, because their published record is too thin to describe a drone programme: flights on fewer than three separate days, or a record that is entirely testing and training. Together they account for 130 flights, about a tenth of one percent of the data. This matters more than the numbers suggest. A department that opened a dashboard, posted three test flights and never returned to it is not a department that has flown three times. Showing that number would not be a small error; it would be a false impression, and a caveat elsewhere on this page would not repair it. Where such an agency turns out to publish properly on another platform, the merge above finds it and the real record appears instead.
- Dashboards whose owner could not be identified are not shown, and a set belonging to the vendors themselves — internal test orgs, employee sandboxes and template dashboards — is excluded deliberately and counted nowhere on this site.
Status labels
- current — a flight was published within the last 60 days. 142 agencies.
- stale — no flight published in more than 60 days. 40 agencies. It may mean the drones stopped flying, or only that the publishing stopped: many agencies upload in batches, weeks apart.
- unreachable — the last refresh could not read the source, so the previous data is shown.
- retired — the source dashboard no longer exists; the last data collected is kept.
The map
The dot for each agency is a single coarse point — where the department operates, not where any drone went. For agencies on Skydio it is the centre of the extent that agency’s own dashboard publishes. For the rest it is the municipality, looked up once through OpenStreetMap’s Nominatim service and then cached, so nothing is geocoded twice. Map data © OpenStreetMap contributors, available under the Open Database Licence. State outlines come from the US Census Bureau via us-atlas.
Refreshing and corrections
This is a monthly snapshot, not a live feed. An agency page is not a picture of what flew last night; it is the shape of a programme over months and years. The date above is when the data was last collected, and each collection is committed to a public repository, so any past version can be reconstructed.
How a refresh behaves depends on the source. Skydio and San Francisco are re-read in full each time, so a flight an agency later removes disappears here too. Flock and Motorola publish only a recent window, so those are additive: once a flight has been collected it is kept, even after the original portal stops showing it. That is the point of collecting them at all.
Spotted something wrong? The most useful thing you can do is compare a figure here against the agency’s own official map, linked from its page, and report the discrepancy.