# Five States Banned the Sale of Precise Location Data. The Edges Still Matter

> Connecticut, Maryland, New Jersey, Oregon, and Virginia now restrict the sale of precise location data. Here is what those laws close, what they leave open, and what you can do about the trail your phone creates.

- **Author:** Kubilay Tunca
- **Published:** 2026-09-01
- **Category:** For Experts
- **Tags:** Location Privacy, Data Brokers, Surveillance, Mobile Security
- **Canonical URL:** https://cyber-security-in-plain-english.com/post/experts/news/five-state-location-sale-ban-has-edges

---

A weather app can know that you are standing in a clinic car park. A discount app can learn that you spend every Thursday evening at a union hall. An advertising library inside an otherwise ordinary game can collect the same kind of coordinates, attach them to an identifier, and send them into a market you have never seen.

Five US states have now drawn a line through one part of that market. On August 31, 2026, the Electronic Frontier Foundation reported that new rules in Connecticut, Maryland, New Jersey, Oregon, and Virginia all ban the sale of precise geolocation data. The laws differ sharply once you ask what else a company may collect, use, keep, or disclose, but the shared rule matters: your movements should not be inventory on an open shelf.

That is progress. It is also easy to overread. A sale ban does not stop every transfer, reach every company, give every person a direct route to court, or remove the copies already scattered through advertising and analytics systems. Your protection still depends on where you live, what a statute calls a sale, which company holds the data, and whether an attorney general has the time and evidence to act.

The useful question is narrower than “Are location data brokers now banned?” They are not. Ask which pipe has been closed, which pipes remain open, and what you can remove from the flow yourself.

## What changed across five states

The common rule is unusually clear by US privacy-law standards. According to [EFF's August 31 comparison](https://www.eff.org/deeplinks/2026/08/privacy-map-part-2-progress-pitfalls-and-fight-enforceable-location-data), Connecticut, Maryland, New Jersey, Oregon, and Virginia now prohibit the sale of precise geolocation data. Their definitions cover location within a specified distance of a person or device rather than protecting a short list of places such as clinics or places of worship.

That broad coverage is the right choice. A narrow protected zone still exposes the journey. If a broker loses sight of a device for the last few hundred metres before a clinic, it can often infer the destination from the route in and the route out. The same problem appears around a journalist's meeting place, an immigration lawyer's office, a protest assembly point, or a domestic-violence shelter. A trail with one small hole remains a trail.

The five laws did not arrive as one coordinated package. Maryland's Online Data Privacy Act came from its 2024 legislative session and imposes a demanding necessity test on sensitive-data processing. Oregon enacted its restriction through HB 2008. Virginia's SB 338 took effect on July 1, 2026, and an [IAPP account of the signing](https://prod.iapp.org/news/a/virginia-governor-signs-legislation-banning-precise-cellular-location-data/) confirms that it amended the state's existing consumer privacy framework. Connecticut's 2026 Public Act 26-64 and New Jersey's A5328 add their own definitions, consent rules, exceptions, and enforcement machinery.

You should resist the neat headline that five states passed the same protection. They reached the same headline rule by different legal routes, then diverged where daily privacy is won or lost.

Maryland goes furthest on data minimisation. Its law says a controller may collect, process, or share sensitive personal data only when that processing is strictly necessary to provide or maintain a specific product or service requested by the consumer. You can read the requirement in the [enrolled Maryland law](https://mgaleg.maryland.gov/2024RS/Chapters_noln/CH_455_sb0541e.pdf). “Strictly necessary” puts the burden on the company to justify the collection. A map needs your position to show the blue dot. A torch app does not need a record of where you slept.

Connecticut takes a consent-led route. Public Act 26-64 requires a clear affirmative act before covered sensitive data is processed, and its definition rejects consent produced through dark patterns. New Jersey also requires consent for covered processing, while Virginia adds narrower protections tied to known children. Oregon's sale ban reaches the market transaction but, as EFF notes in its comparison, adds no general restriction on other processing of precise location data.

Those differences decide what happens before a sale. A company can collect a great deal of data without selling it. It can use the data internally, disclose it under an exception, hand it to a processor, combine it with account records, or retain it until a later rule or business decision changes the route. A strong collection limit prevents the pile from forming. A sale ban addresses one destination after collection has already happened.

The distinction gives you a practical reading rule. When a privacy notice says “we do not sell precise location,” keep reading. Look for collect, process, disclose, share, retain, service provider, advertising partner, measurement, and business purpose. The word “sell” answers one question. Your movements can still travel under other verbs.

## How an ordinary app creates a marketable trail

Precise location rarely enters the market through a screen labelled “Sell my movements.” It begins with a permission prompt. You install a weather app, navigation tool, retail app, dating service, game, or fitness tracker. The app asks for location because one visible feature needs it, because an advertising library wants it, or because nobody removed a default that was convenient during development.

On a modern phone, that location may come from several signals. Satellite navigation can place the device outdoors. Nearby Wi-Fi networks and Bluetooth beacons can refine the estimate. Cell towers keep the network aware of the device at a broader level. The app usually receives a location through the operating system, then decides what to do with it. Turning off one app's permission blocks that app's normal route to the operating-system location service; it does not make the phone disappear from the mobile network.

That limit matters. [EFF's mobile location guide](https://ssd.eff.org/my/module/mobile-phones-location-tracking) separates tracking by mobile towers, operating-system services, apps, and nearby devices. These layers have different observers and different controls. A permission audit can reduce app collection. It cannot prevent a carrier from operating its network or erase location already attached to an account.

Inside the app, third-party code can widen the audience. Software development kits provide advertising, analytics, crash reporting, maps, login, payment, or engagement features. One app may contain several. In July 2026, an [EFF investigation of Android advertising libraries](https://www.eff.org/deeplinks/2026/07/developers-beware-ad-libraries-betray-your-users-location-privacy) found libraries that could collect and share location when the host app held location permission. The user sees one app icon. The data path may involve the app maker, library vendor, advertising exchange, measurement firm, and downstream buyer.

Identifiers make isolated points useful. A coordinate at 8:03 a.m. says little by itself. Repeated points tied to the same advertising identifier, account, device fingerprint, or pseudonymous token can reveal a home at night, a workplace on weekdays, a school run, regular medical visits, religious attendance, and the people whose devices repeatedly appear nearby. A name can be added later by matching the home location or another dataset.

The market rewards accumulation. A location provider can package visits into audiences, offer access through a search tool, score foot traffic, or supply data that another company blends with purchases and demographics. The buyer may never receive a spreadsheet headed with your legal name. Pseudonymous device trails can still identify a person when the pattern points to one home and one workplace.

Federal enforcement has already shown why a company label provides little comfort. In January 2024, the Federal Trade Commission announced an order that would [bar InMarket from selling or licensing precise location data](https://www.ftc.gov/news-events/news/press-releases/2024/01/ftc-order-will-ban-inmarket-selling-precise-consumer-location-data) and require deletion of previously collected data, after alleging that the company had failed to obtain informed consent. The order focused on the path from app collection to audience products. It did not depend on a burglar stealing a database. The ordinary business model was the problem.

This is the mechanism the state sale bans interrupt. They remove legal room for a covered company to turn precise coordinates into a sale under the statute's definition. That can reduce the incentive to collect, because a dataset with fewer lawful buyers is worth less. It can also make contracts and audits simpler: a company can write “no precise-location sales” into its data map and require vendors to follow the same rule.

## What a sale ban closes

The strongest benefit is the plain one. A covered business cannot treat a person's precise movements as a commodity and exchange them through a transaction the law defines as a sale. That blocks a direct route from an app or broker to buyers seeking fine-grained location histories.

It also changes the default argument inside a company. Before a ban, a team might ask whether revenue outweighs reputation risk. After a ban, the legal answer arrives earlier. Precise location must be removed from the product, aggregated until individuals can no longer be picked out, or kept outside the sale entirely. A clean prohibition can do more than a long consent notice because it removes the option rather than trying to explain it.

The all-locations definition deserves equal credit. Sensitive-place lists age badly and expose people at their edges. A legislature will always miss a place that matters to someone: an addiction meeting in a rented room, a source's house, a new clinic, a lawyer's temporary office, or the corner where an informal protest begins. Protecting the data type covers those places without forcing a person to announce why each one is sensitive.

For people under a focused threat, the shift may raise the buyer's cost. A stalker, hostile investigator, extremist group, or poorly supervised official should have fewer commercial doors through which to purchase a ready-made trail. Fewer easy doors matter. An adversary may still compel records, compromise an account, install spyware, follow a person, query another state, or buy a product built around an exception, but each additional step costs time and creates another chance of detection.

That is the right way to measure privacy progress. Invisibility is a sales pitch. Good privacy controls remove cheap routes, separate identities, shorten retention, and make surveillance require more authority or effort. Five sale bans remove one cheap route across a large slice of the commercial market.

National apps may adopt the strongest common rule rather than maintain five state-specific data paths, but the laws do not force that result for every user. Coverage can depend on residency, company size, product type, and statutory exemptions. Your phone will not show a badge saying “protected by the Oregon sale ban.” You still need controls close to the device.

## Where the protection stops

The first edge is vocabulary. Privacy statutes often distinguish a sale from a disclosure to a service provider, a transfer during a merger, a response to a person's request, a security operation, or another listed exception. Some define sale as an exchange for money. Others include “valuable consideration,” which can capture broader bargains. Two companies can move the same data under different legal labels and reach different answers.

Virginia shows why the definition matters. Its consumer privacy framework uses a narrower monetary-consideration concept for sale than statutes that also mention other valuable consideration. A company swapping data for analytics, advertising reach, or another non-cash benefit may therefore require a different analysis. The common headline remains true while the boundary underneath it shifts.

The second edge is collection. Oregon's approach, according to EFF's August comparison, bans sale without adding a wider location-processing limit. A company may have no lawful route to sell a precise trail and still be allowed to collect it for its own covered purposes. That leaves a database available to employees, contractors, legal demands, account compromise, and future policy changes.

Maryland's strict-necessity rule attacks that risk earlier. If a product does not need precise location for the feature the person requested, the company should never form the pile. This is the better architecture. Data that was never collected cannot leak in a breach, appear in an internal search, or become tempting during an acquisition.

Consent creates a different edge. A bright button can produce a legally meaningful choice when the request is specific, informed, and freely given. It can also become a ritual. People trying to open a shop's coupon, check a bus time, or finish account setup will press the button that makes the obstruction disappear. Interface designers know how to make refusal small, grey, repetitive, or costly.

Connecticut directly addresses that trick by excluding dark patterns from consent. The language helps, but a regulator still has to examine the screen and its context. Was the decline button equally visible? Did the service ask again after every launch? Did refusing location remove a feature that did not need location? Did the company describe its partners in words a normal person could understand? Consent quality lives in those details.

Discount programmes open another gap. EFF argues that consent rules in Connecticut, New Jersey, and Virginia retain exceptions that can allow different prices or service levels in some loyalty arrangements. That turns a privacy choice into an economic test. A person with money can decline the tracking. A person counting every grocery bill may accept it for the discount.

The third edge is enforcement. None of the five laws gives every affected person a clear, free-standing right to sue for a location-privacy violation, according to EFF's comparison. Enforcement therefore rests largely with attorneys general or designated agencies. Those offices can bring cases with wide effect, but they have finite staff, competing priorities, and many hidden data flows to investigate.

A violation can remain invisible to the person whose data moved. You may never learn that an advertising identifier linked to your phone appeared in a broker product. Even if you discover it, proving collection, transfer, statutory coverage, and harm can require access to records held by several companies. A right written on paper becomes stronger when the person can obtain evidence and seek a remedy.

The fourth edge is time. A new ban controls conduct from its effective date under its own transition rules. Older copies may remain in backups, derived audiences, or vendor stores unless the law or an enforcement order requires deletion. The FTC's InMarket action included deletion duties because stopping future sales would leave the old stock intact.

A sale ban therefore buys something real and bounded. It closes a commercial route. It does not silence the phone, delete every old coordinate, or stop an adversary with another source of authority. Treating the law as one layer keeps the protection useful instead of mythical.

## Match the response to your threat

Most people do not need to carry a phone in a radio-shielding pouch or abandon navigation. They need to stop granting permanent precise location to apps that can work with less. The aim is proportional control, not a life organised around a theoretical watcher.

Start with the adversary. For ordinary commercial privacy, you are trying to reduce routine collection by apps, ad networks, and brokers. Permission changes, identifier controls, fewer unnecessary apps, and data deletion requests can make a meaningful difference. You can still use maps when you need them.

A survivor dealing with a stalker has a different problem. Account sharing, family-location features, hidden trackers, shared mobile plans, cloud-photo metadata, and physical access to the phone may matter more than a broker sale. Changing settings can alert the other person or remove evidence. A safety plan with a specialist should come before a dramatic reset.

A journalist protecting a source faces another set of links. Repeated co-location can connect two devices even when messages are encrypted. A normal phone carried to a sensitive meeting can create carrier and app records. The right plan may involve leaving identifying devices behind, choosing a meeting method that does not create a unique pattern, and separating research from ordinary accounts. Each measure has a cost and an edge where it fails.

A protest organiser should think about group exposure. One person's location trail can reveal repeated association with others. Photos can carry time and place clues. Ride-share, transit, and payment records may add context. Reducing one app's precision helps, but a serious plan also covers communications, transport, emergency contacts, arrest support, and what participants should do if devices are seized.

Someone facing state surveillance needs to distinguish commercial access from compulsory process. A sale ban can remove a purchase route. It cannot cancel a valid court order, prevent device search after lawful seizure, or stop targeted malware. In June 2026, the US Supreme Court held in *Chatrie v. United States* that accessing historical location records in the case amounted to a Fourth Amendment search; the [Court's opinion](https://www.supremecourt.gov/opinions/25pdf/25-112_0am4.pdf) raises the legal threshold for geofence demands, but it does not erase every lawful route to location evidence.

This separation prevents two bad conclusions. The first says phone settings solve surveillance. They do not. The second says settings are pointless because carriers still know where a connected phone is. They are not. Removing unnecessary app access can eliminate a large commercial branch even when another branch remains.

Think in pipes. Carrier records, operating-system location, app permissions, advertising libraries, account history, photo metadata, nearby trackers, and physical observation are separate routes that sometimes meet. Close the pipes your adversary can afford to use. Then decide whether the remaining routes require stronger operational changes.

## What to do on your own devices

A calm audit takes less than an hour and produces more benefit than installing a mystery “privacy” app. Use the phone's built-in controls first. They sit closest to the data source and do not require handing another company access to your device.

1. **List every app with location access.** On an iPhone, open Settings, Privacy & Security, then Location Services. Apple's [Location Services instructions](https://support.apple.com/en-us/102647) explain the per-app controls. On Android, open the privacy or permission manager and review Location; Google's [location-permission guide](https://support.google.com/accounts/answer/6179507?hl=en) explains approximate and precise choices. Menu names vary by device maker and operating-system version, so use the settings search if needed.

2. **Remove the permissions that have no visible job.** A map needs location while you navigate. A camera may need it only if you want coordinates saved in photos. A calculator, wallpaper app, or simple game has a much harder case to make. Choose “Never” or “Don't allow” where the feature does not depend on place.

3. **Prefer access while using the app.** Permanent background access should be rare. Weather widgets, family safety tools, fitness tracking, and navigation may have a legitimate reason, but you should be able to name it. If you cannot explain why the app needs to know where you are while its screen is closed, remove background access and test the feature.

4. **Use approximate location when precision adds nothing.** A weather forecast usually needs a neighbourhood or town, not a particular doorway. Both iOS and Android expose precision controls for supported apps. Turn precision off, then see whether the service still works. You can restore it for the few features that genuinely break.

5. **Delete apps you do not use.** A denied permission is good; removing abandoned code, accounts, and embedded libraries is better. Check the application account for stored history before deletion. Removing the icon alone may leave cloud records behind.

6. **Review account-level location history and sharing.** Look at family sharing, Find My or equivalent services, map history, photo-location settings, fitness accounts, vehicle apps, ride services, and social check-ins. Device permission and cloud history are separate controls. Clear old history where it has no continuing value, then shorten future retention if the service offers that choice.

7. **Reset or limit advertising identifiers.** This can make it harder to connect new app activity to an older advertising profile, though it will not erase data already matched through an account or another identifier. Also disable personalised advertising where the operating system and major accounts expose the choice.

8. **Check the result after updates.** Apps add features, operating systems change menus, and a permission granted for one trip can become permanent. Put a quarterly reminder in your calendar. A ten-minute repeat audit is enough once the first clean-up is done.

These steps reduce routine app collection. They do not make a powered, connected phone anonymous. Your carrier still needs network information, emergency features may use location under defined conditions, and an app can infer place from an IP address or nearby network even without GPS-grade coordinates. The reduction still matters because precise, persistent trails are far more revealing than a rough city estimate.

For a high-risk situation, change the order. Preserve evidence before resetting a device. Check whether location sharing protects you or exposes you. Use a clean device to seek help if the current one may be monitored. EFF's [Surveillance Self-Defense mobile guidance](https://ssd.eff.org/playlist/privacy-breakdown-of-mobile-phones) provides a sound starting point, and specialist organisations can adapt it to stalking, source protection, protest, or border risks.

## What developers and organisations should change

If your product receives precise location, a privacy notice is the last control, not the first. Start with the data map. Name the screen or feature that requests location, the operating-system permission, the exact precision received, every process that touches it, every vendor that receives it, the retention period, and the deletion path. If one arrow in that map says “analytics” without a vendor, purpose, and field list, keep digging.

Remove collection before polishing consent. A restaurant finder can ask for a postcode or accept a pin the user places manually. A local-news app can default to a chosen town. A delivery service may need a precise address to complete an order but has a weaker case for keeping a continuous background trail after delivery. Product design can often replace surveillance with a direct input.

Treat software libraries as data relationships. Before adding an advertising, analytics, map, or engagement kit, inspect its documentation and network behaviour. Confirm whether it reads location automatically when the host app has permission. Disable optional collection in code. Pin versions, review changes, and test releases by observing what leaves the device.

Contract language should follow the technical map. Prohibit sale of precise location, prohibit use for the vendor's own advertising, name allowed subprocessors, limit retention, require deletion after the service ends, and give your organisation audit rights. A promise from the app means little if the embedded vendor reserves a second purpose in its own terms.

Minimise precision and time. If the feature needs a city, do not store latitude and longitude. If it needs a one-time route, do not retain a year of points. If fraud detection needs a short window, separate that store from marketing and delete on schedule. Aggregation should be tested against re-identification, especially when a small group visits a rare place.

Build refusal as a normal path. The decline button should be as visible as acceptance. The app should explain which feature will be unavailable and leave unrelated features alone. Do not ask again every time the person opens the app. A consent record should preserve the text shown, the version, the choice, and the time so an auditor can examine what happened.

Delete backwards. When policy changes, find historical tables, backups, derived audiences, exports, vendor copies, test data, and cached profiles. Stopping one current API call leaves the old trail in place. The FTC's InMarket order shows why deletion can be part of an effective remedy rather than an optional clean-up.

Finally, test the claim you publish. Capture traffic from a fresh installation, grant and deny permission, move the app through foreground and background states, and inspect what each vendor receives. Repeat the test after library upgrades. “We do not sell precise location” should be an observable property of the system, not a sentence supplied by counsel.

The five state laws give privacy teams a useful minimum rule. The best implementation goes further and makes precise location exceptional, short-lived, and tied to a feature the person knowingly requested. Maryland's strict-necessity language points toward that architecture. You do not need to wait for your own state to copy it.

## The law is a floor you can build on

The August 2026 wave deserves credit. Five states now agree that precise movement trails should not be sold. Their all-locations approach avoids the trap of protecting a clinic doorway while exposing the entire journey. That is a meaningful restriction on a market that has relied on cheap, quiet access to people's lives.

The gaps deserve equal attention. Sale definitions vary. Collection can continue. Consent can be pushed through a manipulative screen. Discount programmes can make privacy expensive. Enforcement agencies may never see the violation, and the person in the trail may have no direct claim.

Your response should match that mixed reality. Use the law where it applies. Ask companies to honour deletion and opt-out rights. Support stronger minimisation, clear private remedies, and bans on charging people for refusing surveillance. On your own devices, remove the permissions and histories that have no job.

The larger lesson is durable: protect the trail before someone names the destination. A location point becomes dangerous through repetition, linkage, and access. Break those links early, keep less, and make anyone who wants the rest spend authority, money, and time.

That is how you become expensive to watch.

For more security and privacy advice in plain English, join the newsletter. One email per month.

## Sources

- [Electronic Frontier Foundation: Privacy on the Map (Part 2): Progress, Pitfalls, and the Fight for Enforceable Location Data Protections](https://www.eff.org/deeplinks/2026/08/privacy-map-part-2-progress-pitfalls-and-fight-enforceable-location-data), accessed 2026-09-01
- [Maryland General Assembly: Chapter 455, Senate Bill 541, Maryland Online Data Privacy Act](https://mgaleg.maryland.gov/2024RS/Chapters_noln/CH_455_sb0541e.pdf), accessed 2026-09-01
- [Virginia Legislative Information System: Senate Bill 338 enacted text](https://lis.blob.core.windows.net/files/1222750.PDF), accessed 2026-09-01
- [IAPP: A View from DC: Kochava is not enough](https://iapp.org/news/a/a-view-from-dc-kochava-is-not-enough), accessed 2026-09-01
- [IAPP: Virginia governor signs legislation banning precise cellular location data](https://prod.iapp.org/news/a/virginia-governor-signs-legislation-banning-precise-cellular-location-data/), accessed 2026-09-01
- [Wiley: New Jersey Adopts Sweeping New Data Broker Law](https://www.wiley.law/alert-New-Jersey-Adopts-Sweeping-New-Data-Broker-Law-Effective-Immediately), accessed 2026-09-01
- [Federal Trade Commission: FTC order will ban InMarket from selling precise consumer location data](https://www.ftc.gov/news-events/news/press-releases/2024/01/ftc-order-will-ban-inmarket-selling-precise-consumer-location-data), accessed 2026-09-01
- [Electronic Frontier Foundation: Developers Beware: Ad Libraries Betray Your Users' Location Privacy](https://www.eff.org/deeplinks/2026/07/developers-beware-ad-libraries-betray-your-users-location-privacy), accessed 2026-09-01
- [Supreme Court of the United States: Chatrie v. United States opinion](https://www.supremecourt.gov/opinions/25pdf/25-112_0am4.pdf), accessed 2026-09-01
- [Apple Support: Turn Location Services and GPS on or off](https://support.apple.com/en-us/102647), accessed 2026-09-01
- [Google Account Help: Manage location permissions for apps](https://support.google.com/accounts/answer/6179507?hl=en), accessed 2026-09-01
- [Electronic Frontier Foundation Surveillance Self-Defense: Mobile Phones: Location Tracking](https://ssd.eff.org/my/module/mobile-phones-location-tracking), accessed 2026-09-01
- [Electronic Frontier Foundation Surveillance Self-Defense: Privacy Breakdown of Mobile Phones](https://ssd.eff.org/playlist/privacy-breakdown-of-mobile-phones), accessed 2026-09-01

---

## About the author

Kubilay Tunca — Senior Full Stack Developer and Author. Founded Cyber Security in Plain English to translate complex security concepts into clear, practical advice, and writes the accompanying books on security, privacy, secure development, and AI systems.

## Books by this author

- **The Digital Fortress** — Your Everyday Guide to a Safer Digital Life. A warm, plain-English guide for people with real lives and finite patience. Learn the handful of habits that genuinely protect your money, accounts, and family, and get honest permission to ignore the rest. [Amazon](https://buy.cyber-security-in-plain-english.com/digital-fortress) · [Details](https://cyber-security-in-plain-english.com/books/the-digital-fortress)
- **The Anonymity Playbook** — Digital Survival for Whistleblowers, Journalists, Activists, and Everyone Else. A practitioner’s field manual for journalists protecting sources, whistleblowers, and activists. It explains how the surveillance actually works, what each technique costs you, and exactly where it fails. [Amazon](https://buy.cyber-security-in-plain-english.com/anonymity-playbook) · [Details](https://cyber-security-in-plain-english.com/books/the-anonymity-playbook)
- **Secure Software Development** — Practical patterns for building secure software. A hands-on security guide for developers and IT professionals who ship real software. Build, deploy, and maintain secure systems without slowing down or drowning in theory. [Amazon](https://buy.cyber-security-in-plain-english.com/secure-software-development) · [Details](https://cyber-security-in-plain-english.com/books/secure-software-development)
- **The Secure Harness** — Shipping Production Code with AI Coding Agents. A calm, practical guide to letting agents do useful work inside boundaries you set, enforce, and audit. Ships with 15 copy-pasteable artifacts: hook scripts, permission configs, release gates, and MCP templates. [Amazon](https://buy.cyber-security-in-plain-english.com/secure-harness) · [Details](https://cyber-security-in-plain-english.com/books/the-secure-harness)
- **The AI Native Engineer** — Build, Evaluate, and Ship AI Systems That Work in Production. Sixteen hands-on chapters, one real product. Grow it from a single model call into a retrieved, tool-using, observable, production-grade system, with evaluation treated as a habit from the first feature. [Amazon](https://buy.cyber-security-in-plain-english.com/ai-native-engineer) · [Details](https://cyber-security-in-plain-english.com/books/the-ai-native-engineer)

Full catalogue with contents and intended audience: https://cyber-security-in-plain-english.com/books

_As an Amazon Associate I earn from qualifying purchases. Buying through these links costs you nothing extra and helps pay for the blog._