id,source,clause,pages,verbalForce,topic,paloBasis,gap,proposedExtension,acceptanceEvidence,priority,status
B-01,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",1; 3,"8,9,10,11,12",scope/definitions,Scope and unwantedness,Risk classification and applicability,"Difference, unwanted bias and unacceptable risk are not identical","Record intended purpose, harms and accountable acceptability separately",Reviewed applicability and scenario rationale,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-02,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",5,"22,23,24,25,26",conceptual model,Lifecycle integration,Six-stage lifecycle,A one-time model test misses operation and retirement,Bind bias studies to all affected lifecycle stages,Versioned activity and reopening records,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-03,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",6.1,"26,27",shall/should distinctions,Profile within the risk file,Case File and evidence bundles,No dedicated executable bias-profile contract,Create a versioned sidecar without changing strict Action Claim fields,"Profile identity, owner, scope and predecessor",P1,Draft analysis / PALO extension proposal; not a conformity finding
B-04,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",6.2,"27,28",shall,Required profile inputs,Data lifecycle and AIMS,"Quality, domain and correctness evidence not composed in one bias protocol","Join purpose, stakeholders, risk, sampling and correctness inputs",Input register with justified missing-input actions,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-05,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",6.3,"28,29",shall,Profile structure and versioning,Evidence references and semantic versioning,Threshold and study versions require stronger binding,"Register groups, methods, scenarios, protocol and changes",Independent protocol registry and digest,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-06,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",6.4,"29,30",shall,Scenario and comparator,Risk scenarios and fairness controls,"Source level, cause and harm may be conflated","Record hazard, group, cause, situation, harm and disjoint comparator",Scenario-level comparison rationale,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-07,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",7.1,"31,32,33",shall; lifecycle integration,Risk-management integration,Governance gates,Numerical results cannot accept residual risk,Separate study findings from authorised risk disposition,Signed reviewed decision and residual-risk record,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-08,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",7.2,"34,35,36",shall,Affected and relevant groups,Subgroup and allocation controls,Attribute inventory alone misses channel and vulnerability,"Derive categories and groups from purpose, evidence and harms",Inclusion/exclusion rationale and sources,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-09,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",7.2,"35,36",shall; conditional justification,Intersections and absence of groups,Subgroup indicators,Small groups could disappear from the assessment,Keep plausible intersectional risks visible and justify any grouping,Coverage inventory including unresolved groups,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-10,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",7.3,"37,38",shall,Hazards across outputs and outcomes,Effect Contracts and harm scenarios,Action correctness does not capture consequence severity,Track agent-stage disparity and downstream harm,Case-level trajectory and harm evidence,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-11,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",7.4.2,"38,39",shall,Per-group quality and correctness,Data fitness and quality-of-service,Scalar or aggregate quality does not establish subgroup fitness,Use the same relevant quality/correctness measures per group,Qualified datasets and group-specific measures,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-12,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",7.4.3,"39,40",shall,Metric selection and custom measures,Indicator registry,Agent metrics need explicit scientific scope,"Justify level, task, hazard coverage and custom names",Clause 11-style metric records,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-13,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",7.4.4,"40,41",shall,False-acceptance family,Research evidence screening,Separate Wilson envelopes do not control a test family,Register the complete iteration family and allocation before estimation,"Protocol, FWER budget and multiplicity calculation",P1,Draft analysis / PALO extension proposal; not a conformity finding
B-14,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",7.4.4,"40,41",shall,Confidence and precision,Measurement methods,Disparity threshold can be confused with precision,Store common confidence and per-estimate margin separately,Interval method and precision results,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-15,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",7.5,"41,42",shall,Reopening triggers,"Incidents, source changes and gates",Router and memory changes can invalidate acceptance,"Trigger review on purpose, version, population, complaint and drift changes",Dependency-linked reassessment events,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-16,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",8.1,43,shall with outcome conflict,Four observation levels,Stack governance,Component results may be mistaken for system findings,Keep data/component/output/outcome evidence separate,Level-specific completion reports,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-17,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",8.2,"43,44",shall,Testing-data qualification,Evaluation lineage,Convenient samples may not represent target groups,"Approve quality, independence and proxy-sampling validity",Dataset snapshot and sampling review,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-18,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",8.2,44,shall; risk-decision route,Data unavailable and privacy,Privacy and access controls,Missing data could be interpreted as zero bias,Return unknown and record the separate risk-management disposition,"Necessity, safeguards, limitations and decision",P1,Draft analysis / PALO extension proposal; not a conformity finding
B-19,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",8.3,"45,46",shall; renamed subclauses,Data bias estimation,Data-fitness evidence,Quality floors and descriptive gaps require distinct checks,Evaluate per-group quality and registered differences,Completion report with all registered groups,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-20,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",8.4,"47,48",should; procedural shall when undertaken,Component analysis,Component evaluation,Optional analysis is not blanket component certification,Apply the full qualified method where component studies are selected,Component test report and context reuse review,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-21,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",8.5,"48,49,50",shall; visible deletions inspected,System output analysis,Action and output verification,End-to-end service outputs may escape model-only evaluation,Evaluate complete system outputs and correctness per group,Output study bound to system version,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-22,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",8.6 heading,"50,51,52",should in visible revision; 8.1 conflict,Outcome analysis,Independent outcome observations,Modal conflict prevents clean requirement assertion,Retain outcome studies as PALO engineering control; resolve source force,Measured consequences and editorial issue record,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-23,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",8; 9,"45,46,48,49,51,52",procedure and reporting,Adaptive best-group comparator,Research bias statistics,Naive pairwise intervals ignore comparator selection,Use a qualified simultaneous/selection-aware design,All candidate groups and comparison selection trace,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-24,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",9,"52,53",shall; tracked edit,Final pre-deployment report,Evidence bundles,Historical publication cannot be inferred as a universal duty,Preserve internal version history; identify required final report,"Current evaluation with counts, intervals and interpretation",P1,Draft analysis / PALO extension proposal; not a conformity finding
B-25,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",9,53,should,Ongoing detection logs,Runtime events and incident packs,A metric threshold is not itself a risk judgement,Link detected events to reviewed criteria and reopening,Threshold rationale and monitoring log,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-26,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",10.3,"55,56,57,58,59,60,61,62,63,64,65",catalogue with per-use requirements,Classification metrics,Allocation and quality-of-service,Rate parity and error parity answer different questions,Choose the beneficial/harmful class and error costs,"Labels, denominators and complementary measures",P1,Draft analysis / PALO extension proposal; not a conformity finding
B-27,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",Regression metrics heading,"65,66,67,68",catalogue with per-use requirements,Mean and loss differences,Latency and burden measures,Equal averages hide distribution tails,"Specify units, loss function, tails and task validity",Regression metric record and uncertainty,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-28,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",Ranking metrics heading,"68,69,70",catalogue with per-use requirements,Ranked representation,Retrieval and allocation,Ranking representation is not observed attention,Register reference shares and task-relevant cutoffs,Ranking protocol and utility trade-offs,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-29,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",Outcome metrics heading,"70,71,72,73",catalogue with per-use requirements,Outcome differences and ratios,Observed effects and remedy,Downstream processes confound causal attribution,Use reliable outcome measurements; reject zero ratio denominators,"Outcome definitions, measurement review and limits",P1,Draft analysis / PALO extension proposal; not a conformity finding
B-30,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",Data metrics: RGI,"74,75,76",shall on purpose and sensitive intermediates,Representation probes,Embedding lineage and privacy,Attribute removal does not remove recoverable proxies,Isolate diagnostic probes; never reuse as production profiling,"Access, purpose, split, retention and probe records",P1,Draft analysis / PALO extension proposal; not a conformity finding
B-31,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",Data metrics: LI,"76,77",catalogue; formula text corrupted,Label imbalance,Annotation and dataset controls,Label proportions are not absolute sample adequacy,"Check labels, counts, prevalence and source interpretation",Metric definition reviewed independently of extraction,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-32,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",10.7[10.8].1,77,shall,Original plus conditioned results,Fairness reporting,Conditioning can hide the initial disparity,Publish both findings and distinct interpretations,Paired original/conditioned report,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-33,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",10.7[10.8].2,78,shall,Conditioning admissibility,Purpose and risk review,Post-hoc proxy conditioning can rationalise harm,Review legitimate features and excluded protected-attribute proxies,Predeclared feature justification,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-34,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",10.7[10.8].3,"78,79",shall; limitations,Stratified comparisons,Intersectional analysis,Sparse cells destabilise findings,Assess every registered cell and retain uncertainty,"Cell counts, confidence and family membership",P1,Draft analysis / PALO extension proposal; not a conformity finding
B-35,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",10.7[10.8].4,"79,80,81",shall/should distinctions,Distribution conditioning,Sampling and comparative evidence,Matching may lack overlap or amplify a few records,Constrain weights/sampling and document balance; no causal claim,"Overlap, sample loss, weight limits and sensitivity",P1,Draft analysis / PALO extension proposal; not a conformity finding
B-36,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",11,"81,82",shall,Metric template completeness,Registry extension design,Agent metrics otherwise lack reproducibility,Record all template elements or justified inapplicability,"Named metric, algorithm, limitations and per-use choices",P1,Draft analysis / PALO extension proposal; not a conformity finding
B-37,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",12.1–12.3,"82,83,84",shall on selection documentation,Treatment design,Control selection,Mitigation may move harm between groups/stages,"Choose data, computation or organisational measures by scenario",Treatment rationale and expected side effects,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-38,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",12.4.1–12.4.4,"84,85,86",method catalogue; external duties separate,"People, interaction and access","Human review, literacy, accessibility",Human involvement can itself introduce disparity,"Qualify UI, review burden and meaningful participation",Representative-user study and accessibility evidence,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-39,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",12.4.5–12.4.9,"86,87,88",method catalogue,Data and model interventions,Data provenance and change control,Synthetic/reweighted data may amplify noise or distort targets,Reevaluate real-world validity and introduced disparities,Before/after datasets and full outcome evaluation,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-40,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",12.4.10–12.4.12,"88,89",method catalogue,"Thresholds, calibration, re-ranking",Router and policy configuration,Better parity can reduce utility or create other harms,Version the intervention and evaluate cross-metric trade-offs,New policy digest and comparative results,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-41,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",12.5–12.6,89,shall,Mitigation evidence and residual risk,Evidence gates and risk ownership,Selected treatment is not proof of resolution,Retest target and other hazards; record residual risk and monitoring,Post-treatment findings and accountable decision,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-42,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",13.1,90,shall; alternative evidence routes,Evidence collection,Source provenance,Expert opinion alone may be incomplete,Record sources and justify expert/research-only selections,State-of-art and impartiality rationale,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-43,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",13.2,"90,91",conditional shall and should,Meaningful consultation,Affected-person and remedy controls,Nominal consultation does not show influence,"Record representation, access, dissent and actual decision use",Consultation record and reuse-context justification,P1,Draft analysis / PALO extension proposal; not a conformity finding
B-44,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",Annex A; Annex B,"92,93,94,95,96,97,98,99,101,102,103",informative,Illustrative groups and bias causes,Risk scenarios and case studies,An illustrative loan case is not a mandatory demographic checklist,Adapt categories to the agentic purpose and lawful evidence,"Scenario-specific relevance, no automatic sensitive inference",Review,Draft analysis / PALO extension proposal; not a conformity finding
B-45,"prEN 18283 supplied N1240 sneak-peek revision, September 2026",Annex ZA,104,informative; conditional future relationship,Limited intended legal coverage,Legal mapping,Draft mapping is not current harmonised conformity,"Map intended Article 10(2)(f)/(g) support, preserving Article 9/14 scope",Source edition and legal applicability decision,Review,Draft analysis / PALO extension proposal; not a conformity finding