HN Debrief

Civic Hygiene – avoid building technologies that could be used by a police state (2013)

  • Privacy
  • Security
  • Regulation
  • Developer Tools
  • Politics

The post is an old Bruce Schneier argument about “civic hygiene.” The idea is not that all technology is evil or that dual-use tools should never exist. It is that some systems are built in a way that makes repression cheap later, especially databases and platforms that collect intimate data, centralize control, or let authorities act at scale with little friction. Commenters spent a lot of time correcting the mangled title, because the useful version of the claim is “don’t build technology that facilitates a police state,” not “avoid anything a police state could ever use.” That narrower framing landed. Most people accepted that cars, computers, and generic infrastructure are too broad a category to reason about. The real target is technology that creates new choke points, new surveillance rails, or new pools of sensitive data that a worse future government can weaponize.

Treat data collection, remote control, and centralized trust as product decisions with long political half-lives. In your own stack, default to data minimization, user control, and architectures that still fail safely when a future government, employer, or acquirer turns hostile.

Discussion mood

Concerned and morally charged. Most commenters agreed with the core warning once it was narrowed from “all dual-use tech” to “systems that lower the cost of surveillance and control,” but they were split on whether the lever is personal ethics, product design, or politics, and many were impatient with naive claims that engineers can stay apolitical.

Key insights

  1. 01

    The title change alters the whole argument

    The strongest clarification was that the post is about technologies that help bring a police state into being, not any tool a police state might touch. That moves the focus away from generic infrastructure and toward systems like registries of sexuality, religion, or immigration status that become turnkey persecution tools when power shifts. This framing rescues the essay from the lazy “everything is dual use” dismissal and turns it into a design question about what kinds of data and control surfaces should never be easy to assemble.

    When reviewing products, ask whether you are creating a new mechanism for population control, not whether the component is theoretically dual use. Sensitive identity data and centralized enforcement rails deserve a much higher bar than general-purpose tooling.

      Attribution:
    • ninkendo #1
    • asveikau #1
    • nonameiguess #1
  2. 02

    Police-state resistance can be a systems design discipline

    One of the most useful technical contributions was a concrete design checklist. Avoid roots of trust users cannot reconfigure. Assume certificate authorities, app stores, telecoms, and your own servers may eventually act against users. Prefer peer-to-peer where possible, publish APIs, allow update refusal, use reproducible builds, and treat binary blobs and hidden dependencies as governance risks, not just security risks. This turns the abstract ethics argument into ordinary architecture work.

    Add political abuse cases to your threat model, especially for products that mediate identity, communications, or updates. If a hostile operator can flip one switch and control everyone, you built the wrong shape of system.

      Attribution:
    • __MatrixMan__ #1
  3. 03

    Telemetry is not neutral product plumbing

    Several comments pushed back on the idea that telemetry is harmless general infrastructure. The argument was that most telemetry begins as opt-out surveillance with weak user consent and only vague product value. Once the data exists, every future owner, manager, and regulator inherits the ability to repurpose it. That makes telemetry less like innocuous instrumentation and more like pre-positioned social control capacity, especially when careers are rewarded for extracting ever more insight and never for collecting less.

    Treat telemetry as a governance decision, not a default engineering practice. Require a concrete user benefit, explicit retention limits, and a justification strong enough to survive an acquisition, subpoena, or policy reversal.

      Attribution:
    • wtallis #1
    • Zambyte #1
    • Xeoncross #1
  4. 04

    Basic heuristics rule out a lot of bad software

    A practical moral shortcut emerged from the comments: do not build systems that interact with people without consent, do not build skinner boxes, let users control their data, and let them control how the tool behaves unless that harms others. That sounds simple, but commenters noted it would eliminate much of mainstream consumer tech as currently built. The point was not purity. It was that the edge cases are not the problem. The market has normalized obviously antisocial design because it pays.

    Use a short ethics checklist at product kickoff and be honest about what it excludes. If the business depends on nonconsensual data capture or manipulative engagement loops, the risk is not theoretical.

      Attribution:
    • __MatrixMan__ #1 #2
    • roughly #1
  5. 05

    Politics here means civic maintenance, not punditry

    The most grounded political argument rejected the idea that engineers can stay above politics by just voting and moving on. The useful definition of politics in this conversation was local and procedural. Know what your city is buying. Submit public-records requests. Join advocacy groups. Watch procurement, retention rules, and data sharing. The point was that police-state capacity is assembled through boring administrative steps long before it looks like a national emergency.

    If your company sells to government or sits on sensitive data, track local procurement and records rules like you track platform changes. Small civic processes are often where dangerous systems become normal.

      Attribution:
    • roughly #1 #2
    • chaps #1
    • preg_match #1
  6. 06

    Ethics education helps with judgment, not virtue

    Comments about CS ethics courses landed on a narrower but more credible value proposition than “teach people to be good.” The better case is that ethics training gives serious people vocabulary, historical examples, and decision frameworks so they can spot bad work sooner and argue against it more effectively. Some also argued software needs stronger professional norms closer to civil engineering, where societal harm is treated as an engineering failure rather than a purely managerial choice.

    Do not expect a one-semester course to fix incentives. Do use ethics training and historical case studies to improve reviews, escalation paths, and the quality of dissent inside engineering teams.

      Attribution:
    • SoftTalker #1
    • asdfman123 #1
    • wnolens #1
    • not_wyoming #1

Against the grain

  1. 01

    Automation can reduce arbitrary human discretion

    A minority view argued that some surveillance or enforcement tech can make society fairer by reducing officer discretion. Speed cameras and red-light cameras were offered as examples where consistent automated enforcement may be preferable to selective human judgment that punishes enemies and excuses friends. The reply did not deny corruption risk. It claimed machine systems at least expose tunable parameters like yellow-light timing and threshold setting that can be constrained by law, while human discretion is harder to bound.

    Do not treat all automation in governance as equivalent. For each system, separate “removes arbitrary discretion” from “enables mass surveillance,” then decide which risk dominates and what parameter controls are possible.

      Attribution:
    • arjie #1 #2
    • goda90 #1
  2. 02

    Authoritarians do not need advanced tech to oppress

    One credible pushback held that the focus on tools can overstate their importance relative to political intent. A determined state can jail, exile, or kill people with or without modern databases and cameras. In that view, technology mostly changes efficiency and speed, not the fundamental outcome, so over-indexing on product restraint can become a comforting distraction from the harder job of stopping authoritarian power itself.

    Do not let privacy-by-design become a substitute for institutional defense. Product choices can raise the cost of abuse, but they will not save you from a state that has already decided to crush dissent.

      Attribution:
    • slg #1
  3. 03

    Data minimization beats moral grandstanding

    Another dissenting view said the article overweights individual conscience and underweights boring governance controls. Shelters, hospitals, and public services often need records to function. The better answer is strict minimization, access control, and retention law, not romantic claims that good engineers will simply refuse to build harmful systems. That critique sharpened the conversation by pushing it from personal purity toward operational policy.

    If you run a system that must hold sensitive data, spend less time on slogans and more on deletion schedules, purpose limits, and access boundaries. Those controls survive staffing changes better than personal ethics does.

      Attribution:
    • rini17 #1

In plain english

app store gatekeeping ↩
A platform owner controlling what software can be installed and under what conditions.
dual use ↩
A technology or tool that can be used for both beneficial and harmful purposes.
Flock cameras ↩
Networked license plate reader and surveillance cameras sold to police and neighborhoods in the United States.
FOIA ↩
Freedom of Information Act, a law that lets people request government records.
peer-to-peer ↩
A system where users connect directly to each other instead of relying on one central server.
reproducible builds ↩
A software build process designed so anyone can recreate the exact same binary from the same source code.
telemetry ↩
Data sent automatically from a remote device back to a receiver or server.

Reference links

Open source and licensing debates

Surveillance and police-state examples

Historical and political context

  • IBM and the Holocaust
    Used to argue that information systems can materially amplify state persecution by making targeting and logistics easier.
  • Simple Sabotage Field Manual
    Invoked as a reminder that refusal and friction can be part of resisting harmful institutions.
  • What Fascism Is
    Shared as background reading on authoritarian political movements in the middle of the police-state discussion.

Specific cited examples and talks