Cities should audit an AI surveillance system as it is actually deployed—not rely on a vendor’s overall accuracy claim or a single benchmark score. A credible audit identifies the system and its purpose, tests the relevant task under realistic conditions, reports false positives and false negatives separately, examines differences across demographic groups where lawful and meaningful, checks the data and operational controls, and evaluates privacy, rights, and public accountability.
Facial recognition has the clearest demographic-performance evidence in the available government material, but “AI surveillance” also includes systems that detect, track, classify, or analyze people, vehicles, and events. There is no single accuracy measure that fits all of them. The city must first establish what its system does, what action follows an output, and who could be harmed by an error.
As an Amazon Associate I earn from qualifying purchases.
What a city audit needs to establish
An audit is not simply a test of whether software can match an image or flag an event. It should connect four questions: Is the system being used for a defined and justified purpose? Does it perform that task under the city’s operating conditions? Are the input data, human decisions, and technical controls reliable? Can affected people and the public understand, challenge, and oversee its use?
Free tools Windows power users keep installed
One-click scans. No signup required.
These questions matter because the same technical error can have different consequences in different settings. A mistaken alert that prompts a brief human review is not equivalent to an erroneous identification that leads to police scrutiny, a denial of service, or another consequential decision. NIST cautions that the impact of an error depends on the application.
#1 Best Overall
- [Ultra-Lightweight Ear Buds Designed for Small Ears] Each earbud weighs only 3.7g and features a compact, ergonomic in-ear design made especially for small ears. Secure, low-profile, and comfortable for workouts, all-day wear or overnight sleeping.The ultra-slim profile sits flush in your ear with no hard edges poking or rubbing, so you can even sleep on your side without discomfort.
- [Immersive Stereo Sound with TOZO OrigX Technology] TOZO OrigX tuning delivers clear vocals, balanced mids, and natural stereo sound for music, podcasts, and videos.
- [Long Battery Life for Daily Use] Get up to 7 hours of playtime on a single charge, with up to 32 hours total using the charging case—ideal for workdays, commuting, and extended listening sessions.
- [Bluetooth 5.3 & Stable Connection] Bluetooth 5.3 provides fast pairing, stable wireless performance, and reduced dropouts as you move around home or office.
- [Deep Bass with Clear Vocals] High-performance drivers produce punchy bass while keeping vocals clean and detailed for everyday listening.
For governance, NIST’s AI Risk Management Framework (AI RMF 1.0) offers a voluntary structure for incorporating trustworthiness considerations into AI design, development, use, and evaluation. NIST describes the framework as voluntary and its framework page says it is being revised. It is guidance, not a substitute for identifying binding law, procurement rules, and local policy that apply to the city’s system. NIST Special Publication 1270 can also inform questions about how bias arises across people, data, and institutional contexts. It is not a city-specific audit mandate. As the publication puts it, “While many organizations seek to utilize this information in a responsible manner, biases remain endemic across technology processes and can lead to harmful impacts regardless of intent.”
Start with an inventory of the deployed system
Before reviewing an accuracy claim, establish what the city has acquired and how it is used. The inventory should cover the complete surveillance process, not just the software component: cameras or other sensors, analytics, databases, contractors, connected systems, and the decisions or investigations that follow an output.
- Ownership and purpose: identify the responsible department, vendor or other third party, stated purpose, approved uses, and the person accountable for the system.
- Deployment: record locations, operating hours or duration, whether the system is live or retrospective, and the people or communities likely to be affected.
- Data: document what is collected, generated, purchased, or matched; where it comes from; who can access it; where it is stored; and with whom it is shared.
- Configuration: record the software and algorithm version, settings and thresholds, connected databases, and any material changes made since procurement.
- Claimed effectiveness: write down the vendor’s stated performance measures and the evidence, conditions, and task those measures refer to.
Portland Resolution 37608 provides a municipal policy example: it directs work on a surveillance-technology inventory that describes technology type, use and purpose, the responsible bureau or third party, and methods for assessing effectiveness. It also calls for annual reporting, oversight strategies, privacy impact assessment procedures, and public and equitable processes. The resolution is an example of a governance approach; it does not establish that every city has implemented these measures or that a particular system passed an audit.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Repair Windows errors before they cause bigger problemsFix Now →Define the task and the consequences of error
“AI surveillance” describes unlike tasks. A system may detect an object, track it across locations, classify an event, verify a claimed identity, or search for a person in a collection of records. The audit question and appropriate test depend on which task is performed and what happens next.
Rank #2
- 【IP68 Waterproof Multifunctinal Smart Watch】Torjalph fitness activity tracker smart watch is IP68 waterproof technology ,you can wear it when swimming,handwashing,it's rain-proof which means that you perform multiple exercise modes outside without worrying about sweat or rain corroding your fitness smart watch,and this watch equipped with a 230 mAh large capacity battery magnetic charging ,and this sports activity tracker watch can be used 5-7 days If fully charged.
- 【Larger 1.83'' Full Screen Watch & Multi-dial Selection】1.83 inch full HD color touch screen smart watch for men / women provides excellent 240*284 resolution image quality ,This Bluetooth smart watch offers a better touch experience and remarkable immersion and ,more than 200 watch faces for you to choose from App"Da Fit ", you can change the dial style according to your mood to wear keeping the world on your wrist
- 【Accurate Activity Tracker Smart Watch 】sports watch activity tracker uses sweat and dust resistant breathable silicone strap, our Fitness Tracker Smart Watch for Android iOS Phones supports 12 sports modes,walking,running,bicycle,swimming,badminton,basketball,football,rope skipping. This bluetooth calls smart watch for men women is lightweight and stylish, bringing more convenient and relaxing sports experience while also providing more protection.
- 【Your 24 Hour Personal Daily Health Monitor】Fitness watch for women men full touch screen IP67 waterproof bluetooth smartwatch for Android iOS Phones with heart rate blood pressure monitor multifunction Andriod smart watch with steps counter sleep monitor sports fitness watch, monitoring blood oxygen saturation allows you to understand your physical condition in real time to track fitness progress ,to know your health status in time.
- 【More Functions Bluetooth Call Smart Watch】This fitness activity smart watch full screen for iPhone and Android phones includes a bunch of new features including:The bluetooth smart watch can hang up the phone, receive and display WhatsAPP /Text /Facebook /Twitter /Messenger reminder.Support sedentary reminder, alarm clock, stopwatch, timer,weather information,music control, remote camera control compatible with Android 5.0 and above or iOS 9.0 and above smartphones.
For facial recognition, distinguish one-to-one verification from one-to-many identification. In one-to-one verification, the system compares two images to assess whether they show the same person. In one-to-many identification, it searches a face against a gallery or watchlist. An error in the second task can create a false lead among many possible candidates, so its risks and interpretation are not interchangeable with a verification result.
| Audit distinction | What it means | Why the city must record it |
|---|---|---|
| False positive | NIST defines this as treating two different people as the same person. | It can cause a person to be wrongly matched, flagged, or subjected to additional scrutiny. |
| False negative | NIST defines this as failing to match two photos of the same person. | It can mean a relevant person or event is missed; the consequence depends on the system’s purpose. |
| One-to-one verification | A comparison of two images to verify a claimed identity. | Test and report it as verification, with its own threshold and operating conditions. |
| One-to-many identification | A search of an image against a gallery or reference collection. | Record gallery size, search conditions, alert handling, and consequences; do not present verification results as evidence of identification performance. |
Before testing, map how an output is handled: who receives an alert, what evidence they review, whether another source is required, and what actions are permitted. Include plausible harms from both errors, such as mistaken accusation, increased surveillance, missed incidents, or denial of a service. This harm model determines which performance measures and safeguards deserve priority.
Measure performance for the actual use
For facial recognition, report false-positive and false-negative performance separately. State the system and algorithm version, task, threshold, test population, image source and quality, environmental conditions, and the procedure used to decide whether a result counts as a match. Where lawful and methodologically appropriate, report results for relevant demographic groups as well as aggregate results. Explain gaps in the test data and avoid treating a small or unrepresentative sample as conclusive.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Do not publish a single “accuracy” percentage without defining what it measures. The term can obscure which error is counted, what threshold was used, whether the test involved verification or identification, and whether the conditions resemble the city’s deployment. A result from one algorithm cannot be generalized to another. NIST summarized this point in its facial-recognition testimony: “Different algorithms perform differently.”
Rank #3
- [Low Noise High Fidelity] New bee guitar cord with 22AWG oxygen-free-copper center conductors provides enhanced signal transfer, the PVC insulation material and the high-density stranded copper double shield provide noise rejection for quiet operation
- [Ideal for Choice] New bee 1/4 straight and right angle metal connectors, ensures a secure connection in any instrument, pedal, or amp jack. L shape great fro on stage guitar stands
- [Wide Compatible] Suitable for mixer, power amplifier, microphone, effector, decoder, equalizer, electoronic organ, electric guitar, electric bass and other electronic equipments. Widely used in the installation of smart home, karaoke speaker, professional stage audio, etc
- [Easy Disassemble] Making incidental repairs quick and easy! Very nice, flexible cable
- [Dont Hesite to Order] 24 hours customer services and professional technology team are standing by
NIST’s 2018 Face Recognition Vendor Test (FRVT), as described on its testimony page, evaluated 127 facial-recognition algorithms from 39 commercial developers and one university using 26 million mugshot images of 12 million individuals. Those figures describe the scope of that evaluation, not the number or makeup of city deployments. NIST’s 2019 demographic-effects evaluation covered 189 algorithms from 99 developers, tested using four photo collections containing 18.27 million images of 8.49 million people. It found demographic differentials, but their size and direction varied by algorithm and application.
In the tested algorithms for one-to-one verification, NIST reported that false-positive rates across demographics often varied by factors of 10 to more than 100, while false-negative differences often varied by factors below 3. These are findings about the algorithms and evaluations studied, not a prediction for every current system or real-world city deployment. The city should use them as a reason to demand task-specific evidence and disaggregated results—not as a shortcut for estimating its own system’s error rates.
For live video, identify the operating conditions that can affect image capture and matching, such as camera placement, lighting, motion, and image quality, and test under conditions that reflect the intended deployment. A NIST-hosted OSAC technical document provides performance metrics and measurement guidance for a described passive live facial-recognition concept of operations. It also addresses proportionality, human rights, privacy, and privacy-by-design. Its scope is that concept of operations, not a universal standard for every surveillance technology.
Audit databases, records, and operational controls
A technically capable algorithm can still produce unreliable results if the reference data are wrong, stale, poorly documented, or handled insecurely. Review the full path from data acquisition through deletion, including what staff do when the system returns a match or other alert.
Rank #4
- Digital Family Calendar: BSIMB digital calendar wall touch screen puts an end to your chaotic life. Designed for busy families, the digital calendar keeps you organised and in step with your family. Assign different colours and tasks to each family member to make family scheduling easier and more intuitive.
- Easy Setup and Auto-Sync: Plug in, connect to WLAN and auto-syn your electric calendar in minutes. Digital wall calendar compatible with Google calendar, iCloud calendar, Outlook calendar, Cozi calendar and Yahoo calendar. Add events directly on the device or through the ecalendar app. Electronic calendar stays current whether you're at work or traveling—no manual updates needed.
- All-In-One Home Assistant: Smart calendar builds habits and encourage independence with personalised chores. Display daily meal plan to reduce the daily ‘What's for dinner?’ stress. To-do lists keep track of everything you want. Turning chores into valuable achievements with stars and rewards, making life easier and more enjoyable.
- Manage Anytime Anywhere: BSIMB electronic calendar for family stay connected with the powerful ecalendar app. Access and manage your digital wall calendar anytime, anywhere with free features like stars & rewards, photo screensaver, chore charts, meals plans, to-do lists etc. Electronic planner paid features like Magic Import tailor advanced tools for users with special needs.
- Stay in Touch with Loved Ones: Share photos or videos from your phone to electronic wall calendar with app, instantly update your life moment, and keep in touch with your family, especially those who are far away; you can also surprise your smart calendar digital wall by sending him a wish card; it's also great for keeping tech-illiterate elders in touch with the youngsters.
- Reference data: establish who can add people, vehicles, or records; the criteria for inclusion; how long entries remain; and how often the database is validated.
- Correction: determine how inaccurate or outdated records are identified and corrected, who can request a correction, and how a corrected record propagates to connected systems.
- Image and record integrity: check image quality and metadata, whether records can be altered without detection, and whether the system preserves a complete audit trail of searches, access, exports, and changes.
- Access and security: identify authorized users, role limits, monitoring, data-sharing arrangements, storage protections, and procedures for responding to misuse or a breach.
- Retention and deletion: document retention periods for source images, alerts, match results, and logs; how deletion is verified; and whether copies held by contractors or recipients are covered.
- Human review: inspect training and written procedures for interpreting results, escalating uncertainty, recording decisions, and preventing a machine-generated alert from being treated as proof.
UK local-authority surveillance guidance covers public-space CCTV, automatic number plate recognition (ANPR), facial recognition, body-worn video, drones, and vehicle-borne video. Its practical questions include whether recognition databases are accurate and up to date, whether entries are regularly reviewed and can be corrected, and whether storage, security, deletion, and audit-trail controls are in place. The GOV.UK page says the guidance was updated August 25, 2026. It is a useful operational checklist, but UK guidance does not determine the legal requirements for a city in another jurisdiction.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Assess privacy, rights, and proportionality separately
Performance evidence cannot by itself establish that surveillance is necessary or acceptable. The city should document the specific problem the system is intended to address, why less intrusive alternatives are insufficient, who is subject to collection, and whether the scale and duration of use are proportionate to the stated purpose.
A privacy and rights assessment should examine data minimization, retention, access restrictions, sharing, misuse risks, security, transparency, and remedies for people affected by an error. It should also consider whether people can contest a decision, correct relevant records, or learn how their information was used, consistent with applicable law. A technically accurate system may still impose disproportionate surveillance or create harms that an accuracy metric cannot capture.
Public participation should happen early enough to affect the decision to acquire or deploy the technology, not only after the city has chosen a vendor. Identify communities most likely to experience the system’s impacts and provide a meaningful way to raise concerns about use, location, safeguards, and alternatives. Portland Resolution 37608 calls for public and equitable processes in developing accountability and oversight strategies, making participation part of surveillance-technology governance rather than an afterthought.
Compare systems on equivalent terms
When evaluating multiple systems or procurement options, compare them on the same task and operating conditions. A vendor’s best-case demonstration should not be compared with another system’s field results. If the available evidence does not cover the city’s intended conditions, record that gap rather than filling it with an assumed value.
| Comparison area | Evidence to request or examine |
|---|---|
| Task and conditions | Same function, deployment setting, threshold, image or sensor quality, and operating assumptions. |
| Errors | False-positive and false-negative results separately, with definitions and denominators. |
| Group performance | Results for relevant demographic groups, test-data limitations, and the method used to make the comparison. |
| Data quality | Source and accuracy of reference data, validation schedule, correction process, and retention rules. |
| Consequences and controls | What follows an alert or missed event, human review procedures, access restrictions, security, and audit trails. |
| Accountability | Independent evaluation evidence, transparency, public participation, and a way to require and verify corrective action. |
NIST’s evaluations show why neither a blanket claim that a product category is accurate nor one that it is inaccurate is adequate: tested algorithms differed, and demographic effects were not identical across systems. The useful comparison is system-specific, tied to the same use case, and explicit about what has not been measured.
Publish findings and make remediation enforceable
A public audit report should give readers enough detail to understand the conclusion without exposing sensitive personal data or creating avoidable security risks. At minimum, include the system and purpose, responsible department and vendors, deployment context, audit method, test population and limitations, error definitions, group-level findings where lawful, data and operational findings, and the privacy and impact assessment.
Separate established findings from unresolved questions. For each problem, name the responsible official or department, the corrective action, a deadline, and how completion will be verified. Possible actions include correcting a database, restricting access, changing an alert-review procedure, repeating a test under realistic conditions, limiting a use, or suspending deployment while a material risk remains. The response should be proportionate to the harm and supported by the city’s authority and policy.
Portland’s policy is a municipal example of mechanisms that can support public accountability: an inventory, effectiveness-assessment methods, annual reporting to city council, privacy impact assessment procedures, and oversight strategies. An audit that reports a problem but assigns no owner, deadline, or follow-up mechanism cannot show whether the city acted on its findings.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
