What’s trending in AI on 10 October 2026: The biggest private surveillance network in the United States is shrinking its workforce. On 9 October Reuters and TechCrunch reported that Flock Safety, whose AI-powered cameras read licence plates for police departments, businesses and neighbourhood associations, is cutting about 18% of its staff, roughly 270 people, through a voluntary buyout that ends this month. The cuts land two weeks after a Senate hearing in which Flock’s CEO declined to testify, and after a year in which city after city switched its cameras off over data sharing, officer misuse and security holes. Below: what happened, what the hearing exposed about camera security, and a scorecard for anyone who runs or shares data from AI cameras.

1. What Flock announced, and what it didn’t

Reuters first reported the cuts, and TechCrunch confirmed them with the company on 9 October. About 18% of Flock’s employees, or roughly 270 people, are taking a buyout and will leave at the end of October. Wired had reported in September that Flock was asking staff to leave voluntarily to avoid mandatory layoffs.

Flock is keen on the distinction. Spokesperson Paris Lewbel told TechCrunch: “This was a voluntary program, not a layoff.” The company would not say whether CEO Garrett Langley or other executives were taking pay cuts.

The framing matters less than the signal. Flock has raised close to $1 billion in venture funding, according to the ACLU, on a pitch of more cameras and more shared data. Companies in that position rarely shed a fifth of their people unless growth has slowed, and the obvious cause is the year-long revolt by the communities the cameras watch.

2. How an AI camera network became a political problem

Automated licence plate readers are not new. Toll roads and police cruisers have used them for about a quarter of a century. What Flock changed was scale and searchability. Its cameras use AI to read plates and make the images searchable, and agencies can search not only their own cameras but, depending on settings, those of other customers. TechCrunch puts the network at more than 100,000 cameras. Flock says it serves more than 4,800 law enforcement agencies and nearly 1,000 businesses.

That reach is exactly what critics object to. Senator Josh Hawley, who chaired the September hearing, argued that the novelty is billions of images becoming searchable, including with AI. Three problems kept turning procurement decisions into council-chamber fights:

• Data travelling further than residents were told. In Mountain View, California, an audit found federal agencies had accessed city data without police knowledge. In Oxnard, an audit found a “nationwide query” setting had shared data with federal agencies, reportedly in breach of state law.
• Misuse by the people with access. TechCrunch, citing 404 Media and USA Today, reports that dozens of officers have been arrested, fired or pushed out for using the system to search for, track or stalk people without a warrant.
• Errors with real consequences. Misread plates have led to improper stops and arrests. At the hearing, Florida resident Lindsey Isaacs said she spent 13 days in jail, accused of vehicular homicide, after camera data placed her car miles from a fatal crash.

Trust also eroded over how the company behaved. Cambridge, Massachusetts ended its contract after cameras were installed without authorisation. In Oshkosh, Wisconsin, the council rescinded a just-approved contract within a day after the police chief said Flock’s description of its heat-map feature conflicted with what his own officers had seen. Once a customer feels misled, the technology’s benefits stop mattering. We saw the same dynamic with AI smart glasses in the workplace and in the new surveillance-pricing laws: people accept AI that watches them only when they know where the data goes.

3. The cancellation wave, counted three ways

There is no official register of cancelled contracts, so the size of the backlash depends on who is counting and what counts. The Washington Examiner tallied about 23 city councils that voted to cancel, not renew, reject or deactivate Flock readers in the first seven months of 2026. The State of Surveillance tracker lists at least 30 cities and agencies since late 2024 and names 19 of them. The ACLU, counting Flock and rival plate-reader vendors together, says more than 200 cities have ended contracts since the start of 2026. None includes private cameras run by businesses and homeowners’ associations.

Denver shows that “cancelled” rarely means “gone”. When its contract expired on 31 March, the city removed all 110 Flock cameras and replaced them with 50 Axon cameras under stricter data controls, while privately owned Flock cameras in the city kept running. Los Angeles police also let their Flock contract lapse, according to TechCrunch.

4. Flock’s August fixes, and why critics weren’t satisfied

Flock did respond. In August it announced a package of policy and product changes, which the ACLU reviewed on 13 August. The table sets out each change next to the main objection to it.

Flock’s change	What it does	The objection
Shorter default retention	Plate data kept for 7 days instead of 30. Flock says about 90% of searches happen within a week.	The ACLU wants 48 hours, and says the next item could keep some data indefinitely.
Evidence Mode	Keeps data tied to a case number beyond 7 days, free to customers.	Depending on scope, it could retain location data on people not suspected of anything.
More local control	A city can decide which offence types other agencies may search its data for, for example blocking immigration searches.	An earlier search-term filter was easily bypassed with vague or joke entries.
Audit Assistance for everyone	A tool that flags abnormal search behaviour becomes standard.	It has never been independently evaluated, so nobody knows how much misuse it misses.
Mandatory case codes	Every search must carry a reason code; this was optional from July 2025.	A dropdown lets users pick any listed reason, accurate or not.

The ACLU called it a response to a PR problem rather than the product’s harms. The tension is commercial: a plate-reader network is more valuable the longer it keeps data and the more widely it shares it, which is exactly what residents fear.

5. The security problem the Senate hearing put on the record

Most of the public debate has been about privacy. The 23 September hearing of the Senate Judiciary subcommittee on Crime and Counterterrorism added a harder-edged concern: a network that records the movements of millions of vehicles is a high-value target, and witnesses argued it is not protected like one. Hawley had invited the CEOs of Flock, Axon, Motorola Solutions and Verkada. None testified, and Langley declined. Hawley instead read out an earlier remark of Langley’s: “We’ve already given up the traditional sense of privacy.”

Cybersecurity engineer Benn Jordan described his 2025 research, as reported by Texas Scorecard. At the time, he said, Flock did not require customers to use multi-factor authentication, and he found Flock law enforcement accounts offered for sale on a dark-web marketplace. He said he found poorly protected data on cameras, including API keys and passwords, and could view, delete and modify footage from 60 of Flock’s newer Condor cameras. Freedom-of-information records from Dunwoody, Georgia, he said, showed Flock employees viewing cameras on that network more than 1,000 times. He added that Flock seems to be taking security more seriously recently.

Alasdair Whitney of the Institute for Justice described a case reported the week before the hearing: thieves took a camera off a pole, copied its storage, recovered an encryption key left on the device and accessed about three weeks of footage showing roughly 50,000 vehicles. He cited Norfolk, Virginia, home to major naval bases, where such a record could interest foreign intelligence. TechCrunch separately notes that some Flock cameras have been found publicly exposed on the web.

None of these failures needs advanced AI to exploit. They are classic device-fleet weaknesses: weak logins, secrets stored on devices and too many insiders with access. We covered the same patterns in the stolen-credentials crisis and in our guide to securing edge and ambient devices. What AI changes is the payoff: a searchable archive of where cars have been is worth far more to an attacker than a single camera feed.

6. Where the rules are heading

Congress has not passed anything yet, but the hearing showed rare cross-party interest. Hawley announced an investigation into Flock’s data practices in August. In the House, Roll Call counts at least four bills. Tim Burchett (R-Tenn.) would bar federal agencies from using plate-reader, facial recognition or biometric tracking. Raja Krishnamoorthi (D-Ill.) and Thomas Massie (R-Ky.) would use federal funding to push states away from the cameras, with exceptions such as tolls. Keith Self (R-Texas) would require a warrant for federal use of local camera data, except in emergencies.

The Institute for Justice is pushing a model state law, which Texas Scorecard describes as the PEEPS Act. It would require a warrant for historical location data, a log of every access, supervisor audits every 30 days, an annual public report and a seven-day retention cap without a warrant. Jordan went further, calling for a federal privacy law comparable to Europe’s GDPR. Even without new laws, the direction is clear: shorter retention, warrants for history, logged access and independent audits. Expect those to become standard due-diligence questions, as we noted in our list of questions to ask AI vendors.

7. What this means if your business runs AI cameras

Flock’s customers are not only police departments. Retailers, campuses, hospitals, logistics yards, property managers and homeowners’ associations run plate readers or AI video analytics, and many share feeds with police. A private operator whose data travelled further than promised would face the same questions the cancelling cities asked, from customers, tenants, employees and regulators. A wrongful stop built on your camera’s data could also land you in a dispute, as we described in our piece on AI evidence in court.

Use the scorecard below to grade any AI camera system you own, are buying or are feeding data into. Score one point per green answer. Seven or more is defensible; four to six needs a review this quarter. Three or fewer means you carry the risk that just cost a well-funded vendor a fifth of its workforce.

#	Question	Green answer (1 point)	Red flag
1	How long is raw plate or face data kept?	7 days or less by default, with longer holds only for named cases	“Whatever the vendor default is” or 30+ days
2	Who outside your organisation can search it?	A written list of named agencies, reviewed every quarter	Nationwide, network-wide or out-of-state sharing left on
3	Does every search need a reason?	Case or incident number required and spot-checked	Free text or a dropdown nobody reviews
4	Are logins protected?	Phishing-resistant MFA, named accounts, no shared logins	Passwords only, or one login per site shared by staff
5	Who audits the access logs?	Someone independent of the users, at least every 30 days	Logs exist but nobody reads them
6	What happens if a camera is stolen?	Encrypted storage, no keys on the device, remote revocation	Nobody knows what is stored on the unit
7	Can the vendor’s staff view your feeds?	Only with your approval, and every view is logged	Routine vendor access with no record
8	Has the system been tested independently?	Third-party penetration test in the last 12 months, findings fixed	Vendor marketing is the only assurance
9	Is a human required before action?	A person verifies any match before a stop, ban or report	Alerts go straight to security or police as fact

Start with the two items behind most 2026 cancellations: turn off any sharing you cannot justify in a sentence, and check whether your contract lets the vendor change defaults or install hardware without written approval. Treat your camera vendor like any third party holding sensitive data; our guide to SaaS supply chain risk covers the contract and access controls that apply here too.

8. What to watch next

• Whether the buyout is enough. Voluntary programmes are often a first step; watch for further restructuring or a pivot towards private-sector customers.
• State bills built on the PEEPS model. Warrant requirements and seven-day caps adopted by even a few states would set a de facto national baseline for vendors.
• Rivals’ positioning. Verkada told Roll Call it has no national network and customers can only access their own data. Expect more vendors to sell “no shared network” as a feature.
• Independent security testing. Mandatory third-party testing of surveillance vendors could arrive through procurement rules before any law does.

Frequently asked questions

Why is Flock Safety cutting jobs?

Flock Safety is reducing its workforce by about 18%, or roughly 270 employees, through a voluntary buyout with departures at the end of October 2026. Flock says it was a voluntary program, not a layoff. The cuts follow a year of cancelled city contracts and criticism over privacy, data sharing and security.

What is Flock Safety?

Flock Safety is a US company that operates a large network of AI-powered automated licence plate reader cameras used by police departments, businesses and neighbourhood associations. TechCrunch puts the network at more than 100,000 cameras, and Flock says it serves more than 4,800 law enforcement agencies and nearly 1,000 businesses.

Why are cities cancelling Flock contracts?

Common reasons include data reaching federal, immigration or out-of-state agencies without local knowledge, officer misuse, misread plates and vendor conduct such as unauthorised installs. Counts vary from about 23 councils (Washington Examiner, January to July) to more than 200 cities across all plate-reader vendors (ACLU).

What changes did Flock make in August 2026?

Flock cut default data retention from 30 days to 7, added an Evidence Mode that keeps case-linked data longer, gave customers more control over which offence types other agencies can search, made its Audit Assistance misuse-flagging tool standard and required a case code for every search. The ACLU said the changes did not go far enough and recommended 48-hour retention.

What security problems were raised at the Senate hearing on Flock?

At the 23 September 2026 hearing, researcher Benn Jordan said Flock did not require multi-factor authentication at the time of his 2025 research, that he found Flock law enforcement accounts for sale on a dark-web market and exposed data on cameras, and could access footage from 60 Condor cameras. Another witness described footage of about 50,000 vehicles copied from a stolen camera.

What should a business that uses licence plate readers do now?

Keep retention to seven days or less, switch off sharing you cannot justify, require a reason for every search, use phishing-resistant MFA, have access logs reviewed independently each month and require a human to verify any match before acting.

---

Sources

• TechCrunch: Surveillance company Flock cuts staff as privacy backlash grows (9 Oct 2026)
• Roll Call: Flock cameras draw bipartisan concerns at Senate hearing (23 Sep 2026)
• Texas Scorecard: Senate hearing examines Flock camera security failures and surveillance concerns (24 Sep 2026)
• ACLU: Despite ‘new’ updates, Flock’s cameras remain a major civil liberties threat (13 Aug 2026)
• Washington Examiner: Flock sees controversial camera contracts canceled (Aug 2026)
• State of Surveillance: 30+ cities have canceled Flock Safety contracts (2026 tracker)