What We Choose to Protect

A new camera appears beside the road. Most people pass it on the way to something else.

The worker is trying to arrive before the shift begins. The parent is thinking about the pickup. The person whose car was stolen last year might welcome the device. Someone else wonders who can search the record of where they have been. Those reactions are not difficult to understand. Each begins with something a person wants protected.

A public argument that recognizes only one of them will be incomplete.

I want a safe neighborhood. I want a person to be able to walk home, keep a vehicle needed for work, and ask for help when threatened. I also want the ordinary movement of people through a community to remain free of unnecessary scrutiny. Safety and liberty both require institutions capable of being questioned.

The harder question is what we build around the camera. Have we invested in the conditions that make life stable? Are services available before someone reaches a crisis? Is the technology effective for the purpose offered in its defense? Who can inspect its use, and what happens when it is used improperly?

These questions concern the kind of society we are financing.

A contract is a public choice

In May 2025, Dallas approved a three-year service contract with Flock Group, covering cameras, installation, subscriptions, maintenance, and support, with an authorized ceiling of about $5.74 million. The item included General Fund money and specified grant funding.¹

That record establishes a decision to authorize spending up to a limit. It does not establish that the entire amount was immediately spent. Nor does it mean that every dollar could simply have been redirected to housing or education. Some funding was restricted to particular purposes. A fair comparison must respect those legal and fiscal conditions.

Once those distinctions are made, the democratic question remains. Why is this the right use of the resources available to this government? What result is expected? What alternatives were considered? How will the public know whether the promised benefit occurred?

The procurement process should be able to answer without treating the questions as hostility toward the police. People who need protection have a strong interest in effective spending. They also have an interest in money not being committed indefinitely to a system whose value is assumed rather than examined.

We should ask the same questions of a housing program, a treatment service, a school initiative, or a transit change. Comparison is not a contest in which one side gets to bring evidence and the other gets to bring good intentions.

But the comparison must exist. A city should not consider the technology in a room from which all the conditions generating emergencies have been excluded.

What the audit actually found

Austin’s city auditor examined the police department’s license-plate-reader program in a May 2025 report. The audit found that APD’s audit process generally complied with council requirements while identifying data and process gaps. Its scan estimate combined Flock and Axon systems, and missing Flock metadata prevented verification of some totals. The report also identified a contractual risk that data could be retained or reused in ways inconsistent with council restrictions.²

That is a documented risk, not proof that the prohibited uses occurred. The audit’s associated arrests are not a causal estimate of arrests that would otherwise have been impossible. These distinctions are essential if we want oversight to improve a real system rather than supply ammunition for a predetermined position.

They also show why oversight must concern more than whether a device works. A camera can read a plate accurately while its records are shared under rules the public did not intend. A local retention promise can be undermined if a contract permits another route for keeping information. A count can look impressive while its meaning remains unclear.

The public needs to know what is being counted. A scan is not a unique person. A record that does not immediately match a wanted-vehicle list is not automatically useless; it might later assist an investigation. It is also not automatically proportionate to retain every such record simply because a future use is imaginable.

That final choice is partly empirical and partly democratic. We can investigate benefits, errors, and costs. We must also decide what limits a free society should impose even when collecting more information might sometimes be convenient.

An unlimited future use is a very large permission to grant for a present-tense promise.

Does the technology help?

It would be comforting for a critic to say there is no evidence that license-plate readers can help. The available research does not justify that blanket dismissal.

An August 2026 working paper by Scott Mourtgos and Ian Adams combines Flock deployment records with crime data from 2017–2023. Its staggered-adoption analysis estimates an 11 percent first-year reduction in vehicle theft after deployment. The study is nonrandomized, and its causal interpretation is sensitive to assumptions about trends. A reported improvement in arrest clearance began before activation, weakening attribution of that improvement to the cameras.³

This is a working paper, not a settled verdict or a verified peer-reviewed finding. It provides evidence of potential benefit that deserves examination. It does not answer every question about costs, privacy, alternative interventions, or which deployment pattern is best.

A different study of intermittent police patrols with vehicle-mounted readers found more stolen-vehicle recoveries but no measured improvement in the assessed deterrence outcomes. Its setting and design differ from a continuous fixed-camera network.⁴

The honest conclusion is that the technology, the deployment, the outcome, and the comparison all matter. Recovering a stolen car is valuable. Preventing a theft is a different outcome. Making an arrest, clearing a case, obtaining a conviction, and improving a resident’s sense of safety are also different. A program should identify which results it seeks rather than move between them whenever one number looks favorable.

We need evidence that can survive a change of sales pitch.

Suppose a city buys a system to reduce theft. If the evaluation later shows a benefit mainly in recovering vehicles, the city may still decide the system is worthwhile. It should explain that revised rationale, report the costs, and allow the public to judge. Accountability does not require every prediction to be perfect. It requires the people making the prediction to remain answerable after the money has been spent.

The cycle we should investigate

Here is the concern at the center of this chapter. A society can make secure participation difficult, allow preventable distress to grow, and then devote increasing attention to managing the visible consequences. Its systems may become skilled at locating people in trouble while remaining poor at helping them establish a stable life.

That is a hypothesis about institutions, not a fact established merely by placing a camera contract beside a homelessness count. To test it in a particular place, we would need a sequence: what services changed, which people were affected, what happened to hardship, how enforcement changed, and what other forces might explain the pattern. We would also need comparable budgets and actual spending, with capital costs, operating costs, restricted funds, and time periods treated consistently.

The work is demanding. The question is still worth asking.

For example, does a person repeatedly come to police attention because an appropriate care response is unavailable? Are officials measuring repeated contacts while failing to measure whether a referral led to help? Is a neighborhood receiving more monitoring while its requests for safer housing or dependable services wait without a decision? Those questions can be investigated through records, service data, and the accounts of residents and workers.

The answer may differ across places. A surveillance purchase may coexist with substantial investment in prevention. A city may face a genuine increase in a specific crime for which the technology is useful. Another may be buying a visible response because the underlying service problem is harder to solve. We need to discover which account fits.

This caution should sharpen criticism. An accusation that cannot survive the budget details is easy to dismiss. A documented pattern of avoidable crisis, failed referrals, and poorly justified expenditure is harder to wave away.

The object is to change the arrangement, not merely to win the description.

Who is being made legible?

There is also an imbalance worth watching in the way institutions gather information. A person can become highly visible as a source of risk while remaining poorly understood as someone trying to live.

The system knows that a vehicle passed a location. Does another part of the system know that the bus needed for an early shift never arrives at the required hour? The first question concerns detection. The second concerns whether participation is possible. Both can matter, but they call for different kinds of attention.

I am not proposing that government build an even more intrusive personal file in order to provide care. That would repeat the problem in a friendlier vocabulary. Much can be learned through aggregate service measures, voluntary feedback, accessible complaints, and careful evaluation that collects only what is needed. A resident should not have to surrender an entire private life to establish that a public service is failing.

The principle is simple: gather enough information to do the legitimate job, establish limits, and make the institution answerable for its choices.

For a surveillance system, those limits should cover purpose, access, retention, sharing, security, and remedies. Who can search? What reason must they record? Who audits the searches? Can data leave the original jurisdiction? What happens after an error or misuse? Can a person challenge a consequential decision? What will cause the authority to expire rather than renew automatically?

The answers should be understandable before the equipment becomes a permanent feature of the landscape.

A promise that only trustworthy people will use a powerful system is not sufficient governance. We build safeguards because personnel, incentives, and political conditions can change.

A fair test before renewal

I would require a public review before a major surveillance contract is renewed. The review should state the original purpose, the actual deployment, the measured outcomes, the uncertainty, and the full cost. It should report breaches of policy and the steps taken in response. It should compare plausible alternatives rather than treating the existing contract as the only available choice.

The people most affected should be able to participate without needing a technical degree or taking an unaffordable day away from work. Police and service workers should contribute their operational knowledge. Civil-liberties concerns should be heard before they become litigation. Residents who have suffered crime should not be used as symbols and then excluded from the details.

Four questions before renewal

1. What problem?

   Defined outcome and baseline

2. What evidence?

   Benefits, uncertainty, full cost, feasible alternatives

3. What limits?

   Purpose, access, retention, sharing, audit, remedies

4. What review?

   Renewal date, public report, conditions to change or end

A proposed accountability framework, not a rating of a particular city.

Figure 2. An original decision framework for public safety spending. The questions apply alongside a comparison with feasible prevention and service investments; this is a proposed accountability process, not a rating of a particular city.

An evaluation should be allowed to produce several outcomes. Continue a program that works within defensible limits. Redesign a deployment whose costs or risks are excessive. End a program whose case no longer holds. Invest elsewhere when a different approach would better serve the public purpose.

This is not hostility to technology. It is what taking technology seriously should look like.

We can apply the same standard to a service intervention without demanding that every human benefit appear in a crime statistic. A housing program has housing outcomes. A treatment program has health outcomes. Some benefits may cross categories, but they should be demonstrated rather than assumed. If a city chooses a service because it regards stable housing as a public good in its own right, it should say so plainly.

We should not make every worthwhile investment disguise itself as policing in order to enter the budget.

Safety that reaches the household

The working person who loses a car can lose much more than a possession. The vehicle may connect them to income, care, and the people who rely on them. A reduction in theft can therefore matter to affordability. Anyone arguing for a more humane economy should recognize that.

The same person may also care about being able to travel without accumulating an unnecessary record available for unrelated uses. They may want a police service that responds when needed and a government that respects the boundaries of the job. These are coherent wishes. We do not need to assign the person to one ideological team before listening.

This brings us back to the larger definition of security. Security includes protection against violence and theft. It also includes protection against a medical bill that cannot be managed, a sudden loss of shelter, or an administrative decision with no effective route of appeal. No institution can provide all of these protections alone. A society has to organize several kinds of competence.

If we invest only in the ability to respond after damage appears, we leave important work undone. If we neglect immediate protection while promising a better future, we also fail people. The challenge is to build the present response and the longer-term foundations together, then check that both are doing their jobs.

There will be tradeoffs. We should name them, make them visible, and revisit them. A community may accept a narrowly defined use of technology while rejecting broader sharing. It may choose a more expensive service because the cheaper one imposes unacceptable risks. It may decide that a program with modest measured benefit does not justify its intrusion.

Those are democratic judgments, informed by evidence rather than replaced by it.

I want a country that is difficult to prey upon and difficult to turn into a collection of suspects. A country that can protect a person on the street and help make the street a place they are passing through on the way to a secure home. A country whose response to vulnerability includes something more ambitious than keeping an accurate record of it.

The camera can tell us that someone passed.

We should care where they were able to go.