D1Steve Kromer
Hot Topics · 01 Surveillance Procurement

Berkeley never put its surveillance contract out to bid.

Not once. Not in 2021, when a Council referral named Flock Safety before any evaluation had happened. Not in 2023, when the contract was awarded. Not in 2025, when the cameras were added. And not in 2026, when the City proposed spending roughly $2 million more with the same company.

Most of the argument about Flock in Berkeley has been about surveillance. That argument is worth having. But it has crowded out a simpler question that the Police Accountability Board and the City Auditor have both now put on the record: did the City follow its own purchasing rules? The answer appears to be no — and that is a problem I can actually fix from a Council seat.

Last updated 17 August 2026

The short version

  1. No competitive bid, ever. Berkeley's Purchasing Manual requires an RFP for service contracts over $50,000. The PAB found no record that the Flock relationship was ever competitively bid at any stage since 2021.
  2. The workaround was a weapons contract. The City "piggybacked" on a cooperative agreement titled Weapons and Threat Detection Equipment, Services, and Other Solutions. License plate readers were added to that contract's price list nine months after it was awarded — so the ALPR pricing Berkeley relies on was never competitively evaluated.
  3. The City Auditor flagged this exact pattern in February 2026 — before the Flock vote — finding 38% of competition waivers unjustified and an average of ~$218,000 in uncompeted expansion per contract amendment.
  4. BPD and the PAB describe the same contract in incompatible terms. One says every City redline was accepted and Berkeley owns its data. The other says the agreement has perpetual licenses, no exit right, and a $75,000 cap that is the City's only remedy.
  5. Crime was already falling. Every category BPD cited dropped in 2025 — robberies 20%, shootings 40%, vehicle thefts 46% — under the existing 52 cameras, before any expansion. The urgency argument was not made on the numbers.
Where I stand

Bid it. Then decide.

I am not going to tell you license plate readers are useless. Los Angeles audited two months of its own system and found 337 vehicle recoveries and 74 felony arrests. That is real. I am also not going to tell you they reduce crime, because fifteen years of independent research says they don't — they help solve some crimes after the fact, which is a smaller and more honest claim than the one being sold to us.

What I will say is this: a city that cannot show its work on a $2 million surveillance purchase has no business asking residents to trust it with the cameras. Verification first. That applies to energy savings, it applies to budgets, and it applies here.

  • Ask 01 Competitive RFP before any renewal or expansion.

    No new contract, extension, or letter of intent with any surveillance vendor until BPD publishes written functional requirements and the City runs a real solicitation. This is the PAB's own recommendation, adopted unanimously on April 22, 2026.

  • Ask 02 Make competitive procurement binding, not advisory.

    Berkeley's Municipal Code does not legally require competition for service contracts — the rules live in a policy manual with no enforcement teeth. The City Auditor recommended closing that gap. I would carry that ordinance. It fixes every future contract, not just this one.

  • Ask 03 Publish the access-log audits. All of them.

    Monthly audits, semiannual published reports, sent to the PAB. Elsewhere, misuse has been found by journalists and litigants rather than by internal review. Assume ours will need outside eyes too, and design for it.

  • Ask 04 Match the terms Oakland and Richmond already got.

    Penalties of $200,000–$290,000 per violation, no carve-out for lookup tools, no federal task-force loophole, and full legal remedies preserved where actual harm exceeds the penalty. Flock accepted these terms next door. There is no reason Berkeley should take less.

  • Ask 05 Report outcomes with denominators.

    "120+ arrests and leads" is not a number anyone can evaluate. Report arrests, charges filed, and cases cleared as separate counts against total alerts and total false alerts. If the system is working, the real numbers will show it.

To be clear about what this is not

This is not a proposal to rip out the 52 existing cameras, and it is not a claim that BPD acted in bad faith. A department under authorized strength reaching for a tool that promises to multiply its capacity is a predictable response to a real staffing problem. The failure here is procedural, it belongs to the City's process rather than to any individual, and process is exactly what a Council seat is for.

The record

Five years, one vendor, no solicitation

This timeline is drawn from the Police Accountability Board's May 4, 2026 memorandum to Council, which reconstructs the procurement history from the City's own staff reports and resolutions. Every item below is traceable to a public document linked at the bottom of this page.

The piggyback, in one sentence

Berkeley bought license plate readers off a cooperative contract for Weapons and Threat Detection Equipment — a contract to which ALPRs were added, by price-list update, nine months after award. Berkeley's own Purchasing Manual permits piggybacking only where the scope of work is identical. And in its bid, Flock took an exception making its own Master Services Agreement the governing document — so the terms that were competitively evaluated are not the terms Berkeley would sign.

Two documents, two stories

Council was handed contradictory accounts of the same contract

Both of these were in front of Council in the same packet. I have not read the Master Services Agreement itself — it was not public — so I am not going to tell you which side is right. I am telling you that nobody voting on May 7 could have known either, and that is reason enough to have waited.

BPD staff presentation · 24 March 2026
Police Accountability Board · 18 March 2026

"City Attorney's Office negotiated MSA accepted in full by Flock — every redline the City proposed was accepted."

"City owns all data, including anonymized derivatives. Flock prohibited from selling, sharing, or distributing City data."

"City's data ownership and control survives contract termination."

"Contractual financial penalties for data breaches or unauthorized disclosures, mirroring framework approved by Oakland Council in December 2025."

"Federal agencies require a court order to access data. Administrative subpoenas are insufficient."

"An irrevocable service license and perpetual anonymized data license that survive contract termination."

"No termination-for-convenience right. Berkeley cannot exit without proving material breach."

"No post-termination data deletion requirement for city-associated data."

"No city consent required before Flock activates new platform features including convoy tracking, predict-a-path, and Flock Nova." No consent required before Flock assigns the agreement to an acquirer.

"$75,000 penalty per violation is sole and exclusive remedy, eliminating legal recourse" — against $200,000–$290,000 that "Flock accepted in Oakland and Richmond." Carve-outs exempt lookup tools and federal task forces.

A separate memorandum from the City Attorney's office, reported in early May, reportedly warned that the City could face litigation exposure and unreliable data protections. I have not seen that memo; it was confidential. That it exists at all is part of the point.

Does it work?

Useful for solving some crimes. Not shown to prevent any.

What the independent research finds. Lum et al. (2011, Virginia) found no crime reduction in ALPR areas versus controls. Koper et al. (2022) found no deterrence difference between ALPR-equipped and ordinary patrol. Koper & Lum (2019) found ALPRs contributed to case clearance in roughly 1% of cases examined, and in nearly three-quarters of the cases where an ALPR was involved it did not contribute meaningfully to resolution. A 2025 review in the Cambridge Journal of Evidence-Based Policing concluded that computing cost-effectiveness is premature. This literature is unusually consistent.

What the tool does do. The LAPD audited its own system for August–September 2025: 498 stolen-vehicle alerts produced 337 recoveries and 74 felony arrests in two months. That is a real investigative return, and I am not interested in pretending otherwise.

And what it costs when it's wrong. The same LAPD audit found roughly one in three hot-list alerts were false — 161 of 498 — caused not by camera error but by stale records: vehicles recovered and never delisted. The Institute for Justice has documented at least 27 cases since 2018 of innocent motorists stopped at gunpoint, detained, or jailed after an ALPR error, and calls that an undercount. That failure mode isn't fixed by buying better cameras. It's fixed by hot-list hygiene nobody writes into a contract.

Berkeley crime, 2024 → 2025
20%↓
Robberies
40%↓
Shootings
31%↓
Commercial burglaries
46%↓
Vehicle thefts
21%↓
Catalytic converter thefts

Source: 2025 BPD Annual Report, as compiled by the Police Accountability Board. Over 2021–2025 the same categories fell 33%, 70%, 13%, 48% and 57%. These declines happened under the existing 52-camera deployment, before any expansion. As the PAB put it: the case for urgency has not been made.

One claim I could not verify

BPD's presentation states that "peer-reviewed studies show 66% better odds of clearance for cases supported by" a real-time crime center. The deck cites no study. There is real research on real-time crime centers and clearance, but I could not trace this specific figure to a source, and a load-bearing statistic in a $2 million ask should come with a citation. If BPD provides one, I will link it here and update this page.

The two risks that don't go away

Sanctuary status and vendor lock-in

Federal access is a legal exposure, not just a value

California's SB 34 bars sharing license plate data with federal or out-of-state agencies. The Attorney General sued El Cajon over it in October 2025. Mountain View discovered federal agencies had accessed its camera data without authorization and terminated its contract; Santa Cruz and Santa Clara County have also walked away. The PAB counts more than 50 jurisdictions nationally that have left Flock over privacy and data-sharing failures. BPD's assurance is a policy governing BPD's conduct — but the data sits on a vendor's national network, and Berkeley does not hold the logs that would reveal a breach. Mountain View had a policy too.

"We're leasing, so there's no lock-in" doesn't survive contact

BPD argued that because the City leases rather than owns the hardware, it can simply not renew. The PAB's answer is the correct one: when a lease ends the cameras go back, but the accumulated scan records, hot-list configurations, alert rules, cross-jurisdictional relationships and staff workflows do not — and migration is controlled by the vendor. Each additional product deepens the dependency until the switching cost itself becomes the argument for staying.

The claim that a single vendor is technically necessary also doesn't hold. Los Angeles runs its Flock data through a different platform entirely. Chula Vista procures cameras and drones separately. Richmond pairs Flock hardware with another vendor's evidence system. A June 2025 Department of Homeland Security market survey evaluated 16 commercially available ALPR systems. Berkeley already contracts with one of the alternatives — Genetec — for parking enforcement.

Look it up

Don't take my word for it

Everything on this page is sourced to a public document. If you think I've read one of them wrong, tell me and I'll correct the page.

Berkeley primary documents

Research and outside reporting

Tell me where I'm wrong More Hot Topics