Axon Is Quietly Becoming the Thing We Fought Flock To Stop

15 min read

A city council I follow voted to kill its Flock contract this spring. Public comment ran two hours. People showed up with printouts of their own tracking logs. The vote passed, the cameras came down, and the local paper ran a photo of a bare pole where the camera used to sit.

Everybody treated that pole as a win.

Six weeks later the same council approved a new vendor, on the consent agenda, no public comment, no photo. The cameras didn’t come back. They went into the streetlights instead, where nobody has to look at them to notice they’re there. Nobody photographs a streetlight.

That vendor was Axon.

TL;DR:

What Actually Happened

If you read the piece on Flock, you already know how the plate-reading business works. What you might not know is that Axon and Flock used to be partners.

Axon was an early investor in Flock Safety. Its patented vehicle recognition tech was baked directly into Flock’s Fleet 3 hardware when it launched. For four years, evidence captured by Flock cameras flowed automatically into Axon’s own Evidence.com platform. They expanded that integration again in August 2024.

Eleven months later, Axon KILLED it.

In July 2025, Axon cut off Flock’s API access, ending a four-year partnership Flock’s CEO Garrett Langley said he found out about as a surprise. He called it a “surprise decision” and said no real reason was given. Axon’s CEO Rick Smith told a different story, accusing Flock of building a “digital silo” to coerce customers into buying more Flock product instead of Axon’s.

Whoever is telling the truth about the breakup, the outcome is not in dispute. Within months, Axon launched two direct competitors to the product it used to feed data into: Axon Outpost, a pole-mounted camera, and Axon Lightpost, built with the exact same streetlight vendor Flock already used.

An early investor turned into the biggest threat to the company it invested in. That’s not an accident. That’s a company that watched Flock build the market, watched the backlash start, and positioned itself to inherit the customers.

Meet Lightpost

Here’s the part that should bother you more than the ALPR backlash did.

Axon Lightpost is a plate reader that clips onto an existing streetlight using a hub built by Ubicquia. No new pole. No trenching. No new power run. Axon’s own materials say it deploys in under an hour. It captures plates at highway speed, does vehicle attribute recognition, and can livestream video, all from a box the size of a shoebox mounted where a streetlight fixture already was.

Compare that to a Flock camera: a visible pole, usually at an intersection or a neighborhood entrance, the kind of thing you notice because it looks like nothing else on the street.

A George Washington University law professor named Andrew Ferguson told NPR in August 2026 that Flock became the center of the surveillance backlash specifically because its cameras are visible. “Once you know what they look like, you become aware of how ubiquitous they are,” was the gist of it. Seeing the camera is what made people start asking questions.

Lightpost’s entire pitch is that you will never have that moment. It disappears into infrastructure that was already there. The surveillance did not get less capable. It got harder to notice. That’s not a side effect of the design. That IS the design.

The Scoreboard

The backlash against Flock is real and it’s working, at least on paper. According to NPR, the crowdsourced tracking site DeFlock shows roughly 100 cities have deactivated cameras, rejected new deals, or cancelled existing Flock contracts.

That’s out of more than 130,000 automated plate readers currently running across the country. So even a full Flock rout barely dents the total surveillance footprint, because the footprint isn’t really about one company.

Rick Smith, Axon’s CEO, said the quiet part on an earnings call: “We’re seeing a significant number of agencies want to come over to Axon and convert from existing license plate readers to us.” Denver signed with Axon. Longmont, Colorado signed with Axon. Syracuse, New York switched, and local critics there are already on record saying the switch solved nothing. Motorola Solutions and Verkada are picking up business too, but Axon is the one positioned to eat Flock’s lunch at scale, because Axon already sells police departments everything else.

This is not communities rejecting surveillance. This is communities rejecting one brand of surveillance and buying a different one from the company that used to be that brand’s biggest partner.

But Axon Says There Is No National Database

This is the claim doing the most work in every Axon pitch, and it deserves a real answer instead of a shrug.

Here’s what’s actually true. Axon’s ALPR sharing model is bilateral, not centralized. Agency A can invite Agency B to search Agency A’s plate data. Agency B has to separately accept. Either side can revoke the relationship at any time. Nothing is copied into a shared pool. Cross-agency searches require a declared offense category, and if the donor agency hasn’t approved that category, the results get silently excluded. On paper, this really is architecturally different from Flock’s National Lookup, where 75% of Flock’s agencies opted into one shared network by default.

Here’s why that distinction matters less than it sounds like it should.

Bilateral agreements compose. Flock’s national network is one switch. Axon’s is hundreds of individual switches that, added together, produce the exact same mesh, except there’s no single dashboard where you can see the whole thing and no single place to file a records request to find out who’s connected to whom.

Axon’s own product documentation includes a workflow literally titled “Establish a sharing partnership with a federal agency.” “No national network” was never a promise that federal agencies can’t get in. It’s a promise about the plumbing, not the destination. In Seattle, a cybersecurity engineer who previously worked with Axon has pointed out that both the Seattle Police Department and ICE hold Axon contracts, and both store their data in the same Evidence.com cloud.

And separately from ALPR entirely, Axon already runs something that functions exactly like a national network: Fusus, which now spans more than 200,000 cameras across the country. It’s just not filed under “license plate reader,” so it doesn’t trip the same alarm.

Here’s the detail that should end the debate. In Axon’s own Fusus documentation, the ALPR Data Sharing Report, the tool meant to let an agency audit who it’s sharing with and on what terms, has two fields called “Device Scope” and “Data Conditions.” Axon’s own help pages describe those fields as hardcoded placeholders, not driven by a relationship’s actual configured settings.

The audit trail that’s supposed to prove the sharing is bounded and accountable is, in Axon’s own words, not actually reflecting the real configuration yet. The receipts for “trust us, it’s controlled” don’t currently show what’s controlled.

None of this means Axon is lying about today’s architecture. It means today’s architecture is a policy choice sitting on top of technology that could centralize with a product update, the same way Flock’s National Lookup was a feature that got added after the cameras were already installed. Nobody voted on Flock becoming a national network either. It just got turned on, and 75% of agencies opted in because opting in was easier than explaining to their own community why they opted out.

Give Denver Its Due

I’m not going to pretend the Axon contracts are identical to what they’re replacing, because the honest version of this argument is stronger than the lazy one.

Denver’s new Axon contract, approved on a narrow 7-6 council vote, replaces 111 Flock cameras with 50 Axon cameras. Data retention drops from 30 days to 21. Searches are limited to Denver Police and Denver Fire arson investigators, and every search requires a case number and a stated reason. The city owns its own data. Axon says it won’t be used to train AI models, won’t run facial recognition against it, and won’t be resold. The whole thing runs $150,000 for twelve months.

Those are real, meaningful improvements over what Flock offered. Denver’s mayor made a point of saying, directly, that unlike Flock, Axon doesn’t operate a searchable national network that could expose the city’s data to agencies in other states.

Now the turn. Every one of those protections is a contract term, not a technical limit. A contract term is exactly the kind of thing that gets loosened at the next renewal, quietly, when nobody’s filling the council chamber anymore because the fight already happened and everyone thinks it’s over. One council member already flagged, in the same meeting that approved these “tighter” terms, that the cameras ship with a livestreaming feature NOBODY asked for.

Retention limits and search restrictions are cheap concessions when the actual business model isn’t the camera. It’s everything the camera plugs into.

Fusus and the Real-Time Crime Center

Axon bought Fusus in 2024. It’s now used by more than 250 policing agencies and connects over 200,000 cameras, a meaningful chunk of them privately owned, into unified police dashboards called real-time crime centers.

The pitch to businesses is voluntary participation. In Rialto, California, a $1.25 million Axon deal wired up more than 100 intersection cameras and pulled in 75 local businesses through a program called Connect Rialto. It got mocked on Last Week Tonight, and city officials pushed back, insisting participation is genuinely opt-in.

Opt-in for the business owner isn’t the same as opt-in for everyone who drives through that parking lot. This is the same pattern from the Flock piece, where Home Depot and Lowe’s parking lots were quietly feeding plate data to police. Fusus is that same idea, generalized, and running under a company that also sells the cameras, the cloud storage, and the report-writing software the cops use afterward.

The Electronic Frontier Foundation has said plainly that Fusus “expands police access to people videoed by private cameras who have not consented to that access.” In Nashville, opponents specifically raised the risk to undocumented residents and to people seeking reproductive or gender-affirming care, groups that have every reason to worry about their movements landing in a searchable police system they never agreed to join.

Draft One Writes Your Police Report

Axon’s other big AI product has nothing to do with cameras on the street. Draft One takes body camera audio and generates the first draft of an officer’s incident report.

The design flaw here is almost elegant in how bad it is. When the officer finishes editing and exports the final report, Draft One deletes the original AI draft, along with any record of which sentences came from the machine and which came from the officer. If an officer’s courtroom testimony ever contradicts something in their written report, they now have a built-in excuse: blame the AI, and there’s no surviving copy to prove otherwise.

This isn’t a hypothetical concern from privacy advocates. It’s already changing policy. King County, Washington’s prosecuting attorney’s office won’t accept AI-assisted police narratives at all. Connecticut’s Chief State’s Attorney imposed a statewide moratorium so agencies can figure out policy before anyone keeps using it. Utah has considered a bill requiring disclosure whenever a report started as an AI draft, because right now there’s no way to tell just by reading the finished version.

A Forbes investigation this July found public records showing the AI gets basic facts wrong, including the names of witnesses and officers. Axon’s response is that a human is always supposed to review and approve the final report. That’s true. It’s also true that Axon’s own internal messages, surfaced in that same investigation, showed officers writing reports at home because they didn’t have time to review them properly at the station, a habit department leadership called “a great example of pure laziness” and asked Axon not to use in marketing.

None of this has slowed the money down. Axon booked roughly $750 million in “AI Era” contracts in the year after Draft One launched, and its AI product revenue grew more than 700% year over year in the first quarter of 2026.

And Now the Drones

Axon’s newest frontier is airspace. It bought Dedrone, a counter-drone and airspace-monitoring company, partnered with Skydio for autonomous “drone as first responder” programs, and added radar tech from a company called Echodyne this past May.

The scale is moving fast. Over 1,000 public safety agencies had received the FAA waivers needed to run automated drone programs as of February 2026, and the agency issued more of those waivers between April 2025 and this February than in the previous seven years combined. Some agencies report their drones clearing up to a quarter of calls without an officer ever showing up.

Here’s the history that makes this land differently than a normal product launch. In June 2022, Axon announced plans to build Taser-equipped drones as a response to mass shootings. Its own AI ethics board had already voted 8 to 4 against the idea. Axon announced it anyway. Nine members of that ethics board resigned in protest, writing that they had “lost faith in Axon’s ability to be a responsible partner.” That same board is the reason Axon banned facial recognition on its body cameras back in 2019. It was, by every account, a real check on the company.

That board doesn’t exist in any functional form anymore. The products it used to have the power to block are now shipping. Axon went on to acquire a military drone manufacturer.

The oversight got resolved by removing the people doing the overseeing.

The Bundle Is the Business Model

Step back from any single product and the shape of the thing becomes obvious. Axon is signing ten-year contracts worth around $45 million that bundle drone programs together with AI report writing, on top of body cameras, plate readers, and the crime center software that ties it all together.

The Electronic Frontier Foundation has been warning about this pattern under the name “Beware the Bundle.” Departments increasingly justify sole-source, no-bid contracts for Axon body cameras specifically because those cameras connect natively to Fusus. The lock-in isn’t a downside of the sales strategy. It is the sales strategy. Once a department buys the camera, every subsequent purchase gets easier to justify and harder to competitively bid, because switching now means losing integration with everything already bought.

Three cities, Baltimore, Augusta in Maine, and Howell in New Jersey, sued Axon in 2024 over exactly this, alleging it has monopolized the body camera and Taser markets. A federal judge in New Jersey tossed the Taser claims in early 2025 but let the body-worn camera monopoly claim move forward. A separate antitrust case over Axon’s acquisition of its old rival VieVu is also still working through the courts.

One company now owns the camera, the cloud it uploads to, the AI that writes the report, the plate reader, the drone, and the dashboard that ties them together. There’s NO meaningful second decision left to make after the first purchase, because every later product is designed to only work well with the first one you already bought.

What To Actually Demand

If the last two years taught us anything, it’s that swapping vendors doesn’t fix a surveillance problem. It just resets the clock on public attention while the actual capability keeps growing. So the fight can’t be “which company.” It has to be policy that survives a vendor switch.

Tier 1: What one resident can push for at the next meeting

  1. Ask for the itemized list of every sharing partnership, federal ones included. It’s a public records request, not a lawsuit.
  2. Ask whether drone policy exists yet, before the department applies for its FAA waiver instead of after the drones are already flying.
  3. Ask why the audit tool has placeholder fields. Make them answer that one on the record.

Tier 2: What takes an organized campaign

  1. Put retention limits in ordinance, not in a contract. Contracts renew quietly. Ordinances require a public vote to change.
  2. Vote on each capability separately. No bundled ten-year contracts that make it impossible to reject the drone program without also losing the body cameras.
  3. No private camera integration without individual, revocable, per-camera consent. Not a business owner’s blanket opt-in on behalf of everyone in the parking lot.

Tier 3: What takes state-level pressure

  1. Require disclosure whenever a report was AI-assisted, and legally require preservation of the original draft. King County and Connecticut got here through prosecutors and moratoriums. Utah tried to get here through statute. Every state needs one of those routes.

None of this requires picking a side between Flock and Axon. It requires treating the vendor as irrelevant to the actual question, which has always been what a community is willing to let any company do with its residents’ movements, on whatever hardware happens to be fashionable that year. For more on what individual privacy hygiene can and can’t fix here, this earlier piece is worth a read, and the broader pattern of a handful of vendors becoming permanent infrastructure inside government is the same story told from a different company’s side in the Palantir piece.

Bottom Line

That council I mentioned at the start genuinely believed it had won something. In a narrow sense, it had. A specific camera, on a specific pole, from a specific company, came down.

But the fight was NEVER actually about Flock. Flock was just the version of this that was easy to see, easy to photograph, easy to point at in a council meeting. Axon studied that fight, learned exactly what made it visible, and built a product whose entire value proposition is that you’ll never get the chance to point at it.

A surveillance system you can’t see isn’t a system that stopped watching. It’s a system that got better at winning.