Quality inspection and defect flags
Inspects products or parts and flags or rejects the defective ones.
How the register minutes it
| Also called | Quality inspection, defect detection |
|---|---|
| Function | Procurement and operations |
| Default stakes | decides alone, when the paste gives none. Units pass or are rejected on the inspection with no person in between; the false negative is a defective unit shipped. |
| Listed use | No Annex III or Article 22 listing |
| Who it faces | Nobody outside the function |
| Where the error lands | The process itself, not a filed statement or return |
| The error that matters | A false negative on a defective unit. The benchmark measures known defects missed on a seeded labelled set, and good units rejected. |
| Review point | A quality engineer audits a sample of passed units and every rejection appeal; the record holds the sample share, the escapes and the period. |
| Oversight record | Who audits the line afterwards, role and competence (Art. 26); date, the tool version and the volume decided; audit sample: the count out of the total, and the share; reversals after audit: count and rate; escalations and to whom; log reference and retention period (Art. 12, Art. 26); input data checked as relevant and representative (Art. 26); incidents reported to the provider and the date (Art. 26). |
| Contract terms to check | Threshold transparency, escape reporting, image retention. |
The benchmark to demand from the vendor
For flags defects in quality inspection, minuted as decides alone: show us performance on 500 of our own labelled cases of quality inspection and defect flags, measured as known defects missed on a seeded labelled set, and good units rejected, at an error rate below 0.5 percent on the sample, or the line moves to decides with review until it is, retested every reporting cycle, and before any change of model or version goes live, with the test set, the metric definitions and the tool version documented so the test can be repeated.
The sample size, the acceptable rate and the cadence follow the stakes class; the register renders this paragraph per line in the vendor benchmark letter.
What each regime attaches
33 clauses across 5 regimesShown on a register for the regimes you tick, on a line minuted at the default stakes; with none ticked, the ISO/IEC 42001 rows are the default. Requirement text in our own words, read clause by clause from the published standards: our statement of each clause, not the instrument verbatim.
The EU AI Act, as deployer
EU AI Act Art. 4 AI literacyProviders and deployers of AI systems must take measures to ensure, to their best extent, a sufficient level of AI literacy among their own staff and any other persons who deal with the operation and use of AI systems on their behalf. The measures must be calibrated to those persons' technical knowledge, experience, education and training, to the context in ...
Common gap: One generic awareness module issued to everyone regardless of role or technical starting point
Source framework: EU AI Act, Regulation (EU) 2024/1689
EU AI Act Art. 5 Prohibited AI practicesProhibits a defined set of AI practices, including subliminal/manipulative techniques causing significant harm, exploitation of vulnerabilities, social scoring by public authorities, predictive policing based solely on profiling, untargeted scraping of facial images, emotion recognition in workplace/education, biometric categorisation inferring sensitive att...
Common gap: Treating exceptions as routine basis
Source framework: EU AI Act, Regulation (EU) 2024/1689
EU AI Act Art. 6 Classification rules for high-risk AI systemsDetermine and record, for each AI system, whether it is high-risk. A system is high-risk where it is intended to be used as a safety component of, or is itself, a product covered by the Union harmonisation legislation listed in Annex I and that product must undergo third-party conformity assessment, or where it falls within an Annex III use case. Where the p...
Common gap: Classification decided once at design time and never revisited when the intended purpose broadened
Source framework: EU AI Act, Regulation (EU) 2024/1689
ISO/IEC 42001:2023
ISO/IEC 42001 A.9.2 Processes for responsible use of AI systemsThe organization shall define and document processes for the responsible use of AI systems, including processes for the use by employees of AI systems provided by third parties.
Source framework: ISO/IEC 42001:2023
ISO/IEC 42001 A.9.3 Objectives for responsible use of AI systemIdentify and record the objectives that steer how AI systems are used responsibly.
Common gap: Are use objectives distinct from development objectives?
Source framework: ISO/IEC 42001:2023
ISO/IEC 42001 A.9.4 Intended use of the AI systemMake sure the AI system is used in line with its intended uses and its accompanying documentation.
Common gap: Is off-label use detected and addressed?
Source framework: ISO/IEC 42001:2023
ISO/IEC 42001 A.10.3 SuppliersSet up a process so that the services, products or materials the organization obtains from suppliers fit its approach to developing and using AI systems responsibly.
Common gap: standard security due diligence used with nothing AI specific
Source framework: ISO/IEC 42001:2023
ISO/IEC 42001 A.6.2.4 AI system verification and validationAI systems shall be verified and validated, and the results documented. Verification confirms requirements are met; validation confirms intended use is achieved.
Common gap: Are V&V results signed off independently from developers?
Source framework: ISO/IEC 42001:2023
ISO/IEC 42001 A.8.2 System documentation and information for usersThe organization shall determine and provide the necessary information for users of the AI system.
Common gap: Are limitations communicated as clearly as capabilities?
Source framework: ISO/IEC 42001:2023
ISO/IEC 42001 A.6.2.6 AI system operation and monitoringAI systems shall be operated and monitored according to organizational and operational requirements throughout their lifetime. Operational guidance shall be available to operators.
Common gap: Is monitoring active and alerts acted upon?
Source framework: ISO/IEC 42001:2023
ISO/IEC 42001 A.6.2.8 AI system event loggingEvent logs shall be generated and recorded during AI system operations to enable monitoring, accountability, and incident investigation.
Common gap: Are logs sufficient to reconstruct an incident or audit a decision?
Source framework: ISO/IEC 42001:2023
ISO/IEC 23894 6.4.2.5 Identification of controlsThe organization should identify the controls already in place that affect the AI risk, including technical measures in the system, organizational measures around it and the human oversight applied, and how effective they are.
Common gap: Existing controls assumed effective without evidence
Source framework: ISO/IEC 23894:2023
ISO/IEC 23894 B.4 Level of automationAI-related risk source: The degree to which the system acts without human involvement determines how far a wrong output can propagate before anyone intervenes, so the level of automation and the human oversight around it are a source of risk.
Common gap: Full automation with no defined intervention point
Source framework: ISO/IEC 23894:2023
ISO/IEC 23894 6.5.2 Selection of risk treatment optionsOptions balance the benefits against the cost, effort and disadvantages of implementing them and are not mutually exclusive: avoiding the risk by not starting or continuing the activity, taking or increasing the risk to pursue an opportunity, removing the risk source, changing the likelihood, changing the consequences, sharing the risk through contracts or i...
Common gap: Every AI risk 'accepted' with no option considered
Source framework: ISO/IEC 23894:2023
ISO/IEC 23894 6.6 Monitoring and reviewMonitoring and review assure and improve the quality and effectiveness of the process and should be a planned part of it, covering planning, gathering and analysing information, recording results and providing feedback (ISO 31000:2018, 6.6); for AI it should determine the effectiveness of the risk management process against the system's actual behaviour in o...
Common gap: No monitoring of the deployed system's behaviour against the assessed risks
Source framework: ISO/IEC 23894:2023
ISO/IEC 42005:2025
ISO/IEC 42005 5.4 Timing of the assessment, reassessment triggers and triagingWhen setting up the process the organization should decide when assessments are carried out and in how much depth, and when an earlier assessment can be used again, adapted to a new purpose or updated, and to what extent. Timing can be shaped by applicable law; contractual and professional duties; the organization's own structures, policies, processes, proce...
Common gap: Assessment done once at design and never revisited
Source framework: ISO/IEC 42005:2025
ISO/IEC 42005 6.3.4 Intended usesThe document should describe the intended uses of the AI system and its intended users (people or information systems), the contexts and conditions of use the system is designed for, and any restricted or sensitive uses identified under 5.7.
Common gap: Intended users undefined
Source framework: ISO/IEC 42005:2025
ISO/IEC 42005 5.8 Performing the AI system impact assessmentThe organization should carry out the assessment by its documented process: find where impacts can come from, the events and their outcomes; weigh how serious and how likely they are; judge them against the thresholds and criteria; choose how to treat them; and record the results in the clause 6 structure. It should take account of the data the system was bu...
Common gap: Assessment reduced to a checklist with no analysis of consequences and likelihood
Source framework: ISO/IEC 42005:2025
ISO/IEC 42005 5.10 Recording and reportingThe organization should record the assessment and its results and report them to the functions and interested parties that need them (other organizational processes, management review, and where appropriate users and other relevant interested parties), retaining the records for a defined period and updating them as needed (ISO/IEC 42001:2023, B.5.3), using t...
Common gap: Results reported nowhere beyond the project team
Source framework: ISO/IEC 42005:2025
ISO/IEC 42005 5.12 Monitoring and reviewThe organization should monitor the AI system and its context after the assessment and review the assessment at planned intervals and on the 5.4 triggers, checking whether identified impacts materialised, whether measures are effective and whether new impacts have emerged, and feeding the results into the management review and the next assessment cycle (the ...
Common gap: Assessment treated as a one-off deliverable
Source framework: ISO/IEC 42005:2025
ISO/IEC 42005 6.7.2 Directly affected interested partiesThe document should identify the individuals and groups directly affected by the AI system, including users, those whose data is processed and those subject to its decisions or outputs, with specific attention to groups with particular protection needs such as children, impaired persons, elderly persons and workers (ISO/IEC 42001:2023, B.5.4; the NIST crossw...
Common gap: Affected parties limited to customers
Source framework: ISO/IEC 42005:2025
ISO/IEC 42005 6.8.2 Benefits and harmsThe document should set out the benefits and harms, actual and reasonably foreseeable, that the AI system brings to individuals, groups and societies in each impact area (fairness, accountability, transparency and explainability, privacy and security, health and safety, financial effects, accessibility, human rights; ISO/IEC 42001:2023, B.5.4 and B.5.5). Tha...
Common gap: Only benefits documented
Source framework: ISO/IEC 42005:2025
ISO/IEC 42005 6.8.3 AI system failures and reasonably foreseeable misuseThe document should describe the impacts of AI system failures (errors, performance degradation, drift, unavailability, adversarial manipulation) and of reasonably foreseeable misuse (6.8.3.3, per the NIST crosswalk), including who is affected, how severely and how likely, and the detection and response arrangements.
Common gap: Failure modes not analysed for impact on people
Source framework: ISO/IEC 42005:2025
ISO/IEC 42005 6.9 Measures to address harms and benefitsThe document should describe the measures adopted or planned to address the identified harms and to realise the benefits: technical and organizational mitigations, human oversight, transparency to affected parties, redress mechanisms, residual impacts accepted and by whom, and how the measures will be monitored under 5.12 (ISO/IEC 42001:2023, B.5.2 treatment...
Common gap: Harms identified with no measures
Source framework: ISO/IEC 42005:2025
The NIST AI RMF 1.0
AI RMF MS-2.3 AI system performance or assurance criteria are measured qualitatively or quantitatively and demonstrated for conditions similar to deployment settings, and measures are documentedAI system performance or assurance criteria are measured qualitatively or quantitatively and demonstrated for conditions similar to deployment setting(s). Measures are documented. Performance is demonstrated under conditions resembling deployment rather than only on held-out test data, and the measures are recorded.
Common gap: Performance demonstrated only in silico on a static test set
Source framework: NIST AI RMF 1.0
AI RMF MS-2.5 The AI system to be deployed is demonstrated to be valid and reliable, and limitations of the generalizability beyond the conditions under which the technology was developed are documentedThe AI system to be deployed is demonstrated to be valid and reliable. Limitations of the generalizability beyond the conditions under which the technology was developed are documented. Validity and reliability are demonstrated before deployment, and the boundary beyond which the demonstration does not carry is written down.
Common gap: Validity claimed from training performance rather than independent validation
Source framework: NIST AI RMF 1.0
AI RMF MN-3.1 AI risks and benefits from third-party resources are regularly monitored, and risk controls are applied and documentedAI risks and benefits from third-party resources are regularly monitored, and risk controls are applied and documented. Third-party data, model, software and hardware dependencies are monitored on an ongoing basis, not assessed once at procurement, and the controls applied are recorded.
Common gap: Third parties assessed at onboarding and not monitored afterwards
Source framework: NIST AI RMF 1.0
AI RMF MP-3.5 Processes for human oversight are defined, assessed, and documented in accordance with organizational policies from the GOVERN functionProcesses for human oversight are defined, assessed, and documented in accordance with organizational policies from GOVERN function. Oversight is designed for the specific configuration, assessed for whether it is exercisable in practice, and consistent with the governing policy rather than invented per project.
Common gap: Oversight designed at a volume the reviewer cannot sustain
Source framework: NIST AI RMF 1.0
AI RMF MN-1.1 A determination is made as to whether the AI system achieves its intended purpose and stated objectives and whether its development or deployment should proceedA determination is made as to whether the AI system achieves its intended purpose and stated objectives and whether its development or deployment should proceed. A recorded determination weighs measured risk against benefit and decides whether to proceed, so that not proceeding remains an available and evidenced outcome.
Common gap: Determination recorded as approval with no evidence weighed
Source framework: NIST AI RMF 1.0
APRA CPS 230
CPS 230 para 28 Design, implement and embed internal controlsThe entity must design, put in place and embed internal controls that reduce its operational risks consistently with its risk appetite and that allow it to meet its compliance obligations.
Common gap: Controls documented but not linked to risks or obligations
Source framework: APRA CPS 230
CPS 230 para 48 Register of material service providers and management of their risksThe entity must identify its material service providers, keep a register of them and manage the material risks of using them. A material service provider (or material arrangement) is one the entity relies on to carry out a critical operation or one that exposes it to material operational risk; it may be an unrelated third party, a related party or a connecte...
Common gap: Intra-group providers left off the register
Source framework: APRA CPS 230
CPS 230 para 55 Manage the risks of each material arrangement, including BCP execution and orderly exitFor each material arrangement the entity must (a) identify and control risks to the provider's ability to keep delivering the service; (b) identify and manage risks the arrangement creates for the entity itself, such as contagion or step-in risk; (c) make sure it can carry out its BCP if required; and (d) make sure it can exit the arrangement in an orderly w...
Common gap: Exit plans exist on paper only, never tested for feasibility
Source framework: APRA CPS 230
CPS 230 para 35 Minimum set of critical operations unless justifiedUnless it can justify a different view, the entity must at least treat these as critical operations: for an ADI, deposit-taking and management, payments, settlements, clearing and custody; for a general, life or private health insurer, claims processing; for an RSE licensee, fund administration and investment management; and for every regulated entity, handl...
Common gap: Customer enquiries omitted from the register
Source framework: APRA CPS 230
Findings this task type can raise
- No benchmark evidence on the entity's own cases
Every classified line where the paste records no benchmark on the entity's own labelled cases. A general benchmark says what the tool does on somebody else's data; the regimes ask for performance demonstrated under conditions like the deployment. The finding is the letter: per line, the sample size, the error type, the rate the stakes class implies and the retest cadence tied to the reporting cycle. - Decides alone, or decides a listed use, with no impact assessment recorded
A line that decides alone, or decides with review on a listed use, with no impact assessment recorded in the paste. ISO/IEC 42005 asks for the assessment before deployment and on the reassessment triggers, ISO/IEC 42001 A.5.2 for the process behind it, and the GDPR for a DPIA where a solely automated decision is evaluated. The assessment may exist; the register cannot show it. - Concentration on one tool
One tool behind a third or more of the lines, or one tool behind every line that decides alone. Concentration is often the right operational choice; the regimes ask that the dependency be recorded, that the supplier be assessed and monitored, and that the entity can keep deciding if the tool is unavailable or withdrawn. - Log retention not evidenced
A line that decides, with the EU AI Act ticked and no log retention recorded in the paste. The deployer keeps the automatically generated logs for at least six months; ISO/IEC 42001 A.6.2.8 asks for event logs that support accountability and investigation. The logs may exist; the register cannot show them.
Do this for every line on your list
Paste the list and get this minute for every task at once, with the stakes class, the listed-use flags, the findings and the obligation rows per regime. Eight lines free, no account.
Build my decision register