open/ms-ai-architect
1
0
Fork 0
Code Issues Pull requests Releases Activity

feat(ms-ai-architect): R7.4 bølge 3 — payload-20..29 dømt+ingestet (245 claims, 10 flagged/0 pass, 72 flagg; ledger 162→172) [skip-docs]

Browse source 
Alle 10 filene i responsible-ai/ dømt av separate Opus-agenter med live
MS Learn-fetch. Ingen pass-record, så re-judge-gaten utløses ikke og
apply-verified-stamp.mjs er ikke kjørt (verified står uendret på 12).

validate-result.mjs: 10/10 OK (slot-binding, claim-id-mengde, lukket
verdikt-sett, folding-heuristikk). Suite 947/947 grønn.

R7.4: 29/49 dømt. Flagg totalt 585 → 657.
This commit is contained in:
Kjell Tore Guttormsen 2026-07-31 18:30:16 +02:00
commit e017f8bcc2
1 changed files with 975 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

976
scripts/kb-eval/data/judge-pass-manifest.json
View file

@ -4,7 +4,7 @@
"derived_from": "scripts/kb-update/data/full-pass-worklist.json (243 due, all never-verified)",
"cadence": "R7–R10 (5 økter × ~49, 8–10 samtidige)",
"batch": "R7.1",
"count": 162,
"count": 172,
"generated": "2026-07-18"
},
"files": [
@ -8242,6 +8242,220 @@
}
]
},
{
"file": "skills/ms-ai-governance/references/responsible-ai/algorithmic-accountability-auditability.md",
"batch": "R7.4",
"judged_at": "2026-07-31",
"per_file_verdict": "flagged",
"claim_count": 23,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-governance/responsible-ai/algorithmic-accountability-auditability.md#4",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/cloud-adoption-framework/ai-agents/governance-security-across-organization",
"evidence_quote": "The second layer, \"Agent observability,\" contains Microsoft Agent 365, Microsoft Defender for Cloud, Azure Log Analytics, Application Insights, and Cost Management.",
"reason": "Application Insights, trace-visning i portalen og Log Analytics er dokumentert, men Agent Identity står i kildens identitets-/kontrollplanlag og finnes ikke i noen observability-oppregning — én bærende del av firedelingen holder ikke.",
"file": "skills/ms-ai-governance/references/responsible-ai/algorithmic-accountability-auditability.md",
"line": 60,
"claim": "Observability-komponentene i Foundry består av: Application Insights (samler traces, spans og telemetri) | Trace Viewer i Foundry Portal (execution timeline, input/output-data, performance metrics, error details) | Agent Identity (Microsoft Entra Agent Identity) | Centralized Logging (Azure Log Analytics Workspace).",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/algorithmic-accountability-auditability.md#5",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/cloud-adoption-framework/ai-agents/governance-security-across-organization",
"evidence_quote": "Require each agent to operate under a distinct agent identity. Use Microsoft Entra Agent ID to assign identity, permissions, and lifecycle controls.",
"reason": "Unik identitet, ownership og lifecycle bekreftes, men verken «versjon» eller et skille mellom production-, development- og test-agenter finnes i kildens oppregning (registeret sporer ownership, purpose, platform og access scope; registerfiltrene er active/blocked/reassigned).",
"file": "skills/ms-ai-governance/references/responsible-ai/algorithmic-accountability-auditability.md",
"line": 62,
"claim": "Microsoft Entra Agent Identity gir hver agent en unik identitet med ownership | versjon | lifecycle status, og skiller mellom production-, development- og test-agenter.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/algorithmic-accountability-auditability.md#14",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/en-us/purview/audit-copilot",
"evidence_quote": "For user interactions with Copilot or Cowork, this property contains CopilotInteraction. For user interactions with other AI applications, values like ConnectedAIAppInteraction and AIAppInteraction are used, as described in the RecordType row.",
"reason": "Den kanoniske oppregningen av AI-aktiviteter i Purview Audit har CopilotInteraction, ConnectedAIAppInteraction og AIAppInteraction — verken AIServiceUsed, PromptSubmitted eller ResponseGenerated finnes, så tre av fire navngitte aktiviteter er fraværende.",
"file": "skills/ms-ai-governance/references/responsible-ai/algorithmic-accountability-auditability.md",
"line": 321,
"claim": "I Microsoft Purview Audit kan man søke på AI-aktivitetene CopilotInteraction | AIServiceUsed | PromptSubmitted | ResponseGenerated.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/algorithmic-accountability-auditability.md#15",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/microsoft-copilot-studio/advanced-bot-framework-composer-capture-telemetry",
"evidence_quote": "Go to the Settings page for your agent, and select Advanced. Within the Application Insights section, enter the Connection string.",
"reason": "Integrasjonen med Application Insights finnes, men aktiveres under Settings → Advanced → Application Insights (der «Enable logging» er en underinnstilling), ikke under «Settings → Logging» slik claimen oppgir.",
"file": "skills/ms-ai-governance/references/responsible-ai/algorithmic-accountability-auditability.md",
"line": 337,
"claim": "Copilot Studio støtter integrasjon med Azure Application Insights for sentralisert telemetri, aktivert under Settings → Logging.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/algorithmic-accountability-auditability.md#22",
"judge_verdict": "not_grounded",
"rule": "R4",
"evidence_url": "https://learn.microsoft.com/en-us/purview/audit-solutions-overview",
"evidence_quote": "The default retention period for Audit (Standard) changed from 90 days to 180 days. Audit (Standard) logs generated before October 17, 2023, are retained for 90 days. Audit (Standard) logs generated on or after October 17, 2023, follow the new default retention of 180 days.",
"reason": "E3 gir Audit (Standard), men gjeldende standard-retention er 180 dager; 90 dager er den historiske raden fra før 17. oktober 2023, så tallet i claimen er utdatert.",
"file": "skills/ms-ai-governance/references/responsible-ai/algorithmic-accountability-auditability.md",
"line": 397,
"claim": "Microsoft 365 E3 uten Purview Compliance add-on gir kun basic audit log med 90 dagers retention og begrenset eDiscovery.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/algorithmic-accountability-auditability.md#23",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/azure/machine-learning/concept-plan-manage-cost?view=azureml-api-2",
"evidence_quote": "",
"reason": "Verken kostnadssiden for Azure Machine Learning eller MLOps-siden sier noe om at Model Registry og Lineage Tracking er uten ekstra kostnad; kostnadssiden lister bare medfølgende ressurser (Blob Storage, Container Registry, Key Vault, Azure Monitor) og compute — ingen Learn-side bekrefter eller motsier prispåstanden.",
"file": "skills/ms-ai-governance/references/responsible-ai/algorithmic-accountability-auditability.md",
"line": 407,
"claim": "Model Registry og Lineage Tracking i Azure Machine Learning har ingen ekstra kostnad — de er inkludert i Azure ML workspace.",
"disposition": "unsourced"
}
]
},
{
"file": "skills/ms-ai-governance/references/responsible-ai/bias-detection-mitigation-strategies.md",
"batch": "R7.4",
"judged_at": "2026-07-31",
"per_file_verdict": "flagged",
"claim_count": 30,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-governance/responsible-ai/bias-detection-mitigation-strategies.md#10",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/azure/foundry-classic/concepts/model-catalog-content-safety",
"evidence_quote": "For image models, the default content filtering configuration is set at the low configuration threshold, filtering at this level or higher.",
"reason": "Medium-terskelen gjelder tekstmodeller, men claimens andre bærende del - samme Medium-terskel for image models - motsies: bildemodeller har Low som default.",
"file": "skills/ms-ai-governance/references/responsible-ai/bias-detection-mitigation-strategies.md",
"line": 121,
"claim": "Default-konfigurasjonen i Azure AI Content Safety er Medium severity threshold (blokkerer medium og høyere) for både text models og image models — samme terskel for alle fire kategorier.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/bias-detection-mitigation-strategies.md#14",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/azure/machine-learning/how-to-responsible-ai-insights-sdk-cli?view=azureml-api-2",
"evidence_quote": "microsoft_azureml_rai_tabular_insight_constructor ... microsoft_azureml_rai_tabular_causal ... microsoft_azureml_rai_tabular_counterfactual ... microsoft_azureml_rai_tabular_erroranalysis ... microsoft_azureml_rai_tabular_explanation ... microsoft_azureml_rai_tabular_insight_gather",
"reason": "Den kanoniske komponentsiden lister alle RAI-komponentene i azureml-registeret, og microsoft_azureml_rai_tabular_fairness er ikke blant dem - fairness beregnes av selve konstruktør-/model overview-komponenten.",
"file": "skills/ms-ai-governance/references/responsible-ai/bias-detection-mitigation-strategies.md",
"line": 194,
"claim": "Azure ML-komponentregisteret inneholder RAI-fairness-komponenten `microsoft_azureml_rai_tabular_fairness`, som hentes med label «latest».",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/bias-detection-mitigation-strategies.md#17",
"judge_verdict": "not_grounded",
"rule": "R7",
"evidence_url": "https://learn.microsoft.com/azure/databricks/data-governance/unity-catalog/data-quality-monitoring/data-profiling/create-monitor-api",
"evidence_quote": "For information about the deprecated quality_monitors API, see Create a data profile using the quality_monitors API (deprecated).",
"reason": "Claimens bærende SDK-identifikator DataQualityMonitor.create finnes ikke: gjeldende API er w.data_quality (create-monitor), og den gamle var quality_monitors - metrikkdelen stemmer, men navnet gjør at en leser ville kopiert feil kall.",
"file": "skills/ms-ai-governance/references/responsible-ai/bias-detection-mitigation-strategies.md",
"line": 303,
"claim": "Databricks tilbyr DataQualityMonitor.create for fairness-/bias-overvåkning av inference-logger med parametrene table_name, inference_log, problem_type og slicing_exprs, og beregner automatisk metrikkene predictive_parity, predictive_equality og equal_opportunity.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/bias-detection-mitigation-strategies.md#18",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/azure/machine-learning/how-to-responsible-ai-insights-sdk-cli?view=azureml-api-2",
"evidence_quote": "rai_causal_component = ml_client_registry.components.get(name=microsoft_azureml_rai_tabular_causal, label=latest)",
"reason": "Den kanoniske SDK-siden dokumenterer komponentkall via ml_client_registry.components.get, og nevner RAIInsights kun som objektet i dashboard-porten; ingen Learn-side navngir metodene RAIInsights.from_model, add_fairness, add_error_analysis, add_explainer, add_counterfactual eller add_causal.",
"file": "skills/ms-ai-governance/references/responsible-ai/bias-detection-mitigation-strategies.md",
"line": 409,
"claim": "Responsible AI Dashboard-komponentene har SDK-metodene RAIInsights.from_model() (oppretter dashboard) | add_fairness() | add_error_analysis() | add_explainer() | add_counterfactual() | add_causal().",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/bias-detection-mitigation-strategies.md#20",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/ai-services/content-safety/overview",
"evidence_quote": "In Content Safety Studio, the following Azure AI Content Safety features are available: Moderate Text Content ... Moderate Image Content ... Monitor Online Activity",
"reason": "Preview-statusen stemmer, men den andre bærende delen gjør det ikke: Content Safety Studios dokumenterte funksjonsliste inneholder ikke custom categories, og kvikkstarten trener dem via REST-API eller Azure AI Foundry sin Guardrails + controls-fane.",
"file": "skills/ms-ai-governance/references/responsible-ai/bias-detection-mitigation-strategies.md",
"line": 503,
"claim": "Custom categories i Azure AI Content Safety er i preview og trenes via Content Safety Studio.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/bias-detection-mitigation-strategies.md#22",
"judge_verdict": "not_grounded",
"rule": "R3",
"evidence_url": "https://learn.microsoft.com/ai-builder/custom-overview",
"evidence_quote": "Document processing: Extract custom information from documents. Category classification: Classify texts into custom categories. Entity extraction ... Prediction ... Object detection ... Azure machine learning models",
"reason": "To av de fire navnene er erstattet i gjeldende taksonomi: Form Processing heter nå Document processing og Text Classification heter Category classification.",
"file": "skills/ms-ai-governance/references/responsible-ai/bias-detection-mitigation-strategies.md",
"line": 575,
"claim": "AI Builder-modelltypene omfatter Form Processing | Text Classification | Prediction | Object Detection.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/bias-detection-mitigation-strategies.md#24",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/azure/machine-learning/concept-responsible-ai-dashboard",
"evidence_quote": "",
"reason": "Siden omtaler Fairlearn, InterpretML, DiCE og EconML som open source-pakker, men ingen Learn-side oppgir lisensvilkåret (ingen tilleggslisens i Azure ML) eller Apache 2.0 for Fairlearn.",
"file": "skills/ms-ai-governance/references/responsible-ai/bias-detection-mitigation-strategies.md",
"line": 726,
"claim": "Responsible AI Dashboard er inkludert i Azure ML-lisensen uten tilleggslisens, og Fairlearn (Apache 2.0), InterpretML, EconML og DiCE er alle open source.",
"disposition": "unsourced"
},
{
"id": "ms-ai-governance/responsible-ai/bias-detection-mitigation-strategies.md#25",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/azure/ai-services/content-safety/overview",
"evidence_quote": "",
"reason": "Learn oppgir kun at tjenesten har F0- og S0-prisnivå og at bruk kan faktureres separat; selve prismodellen (pay-per-use uten base fee) står bare på den JS-rendrede prissiden.",
"file": "skills/ms-ai-governance/references/responsible-ai/bias-detection-mitigation-strategies.md",
"line": 731,
"claim": "Azure AI Content Safety krever egen lisens etter pay-per-use uten base fee, og faktureres separat fra Azure OpenAI/model inference.",
"disposition": "unsourced"
},
{
"id": "ms-ai-governance/responsible-ai/bias-detection-mitigation-strategies.md#26",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/microsoft-copilot-studio/guidance/responsible-ai",
"evidence_quote": "",
"reason": "Den siterte siden dekker ansvarlig AI-prinsipper for Copilot Studio, men sier ingenting om lisensinkludering eller fravær av per-use-kostnad for bias detection.",
"file": "skills/ms-ai-governance/references/responsible-ai/bias-detection-mitigation-strategies.md",
"line": 735,
"claim": "Responsible AI-funksjoner er inkludert i standard Copilot Studio-lisens, uten per-use-kostnad for bias detection-funksjoner.",
"disposition": "unsourced"
},
{
"id": "ms-ai-governance/responsible-ai/bias-detection-mitigation-strategies.md#27",
"judge_verdict": "not_grounded",
"rule": "",
"evidence_url": "https://learn.microsoft.com/azure/machine-learning/how-to-responsible-ai-insights-ui?view=azureml-api-2",
"evidence_quote": "Select Create Responsible AI dashboard (preview).",
"reason": "Gjeldende dokumentasjon merker Responsible AI dashboard som public preview (også i AutoML-artikkelen), så statusen GA motsies.",
"file": "skills/ms-ai-governance/references/responsible-ai/bias-detection-mitigation-strategies.md",
"line": 939,
"claim": "Responsible AI dashboard (Azure Machine Learning) har status GA.",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-governance/references/responsible-ai/content-safety-implementation.md",
"batch": "R7.1",
@ -8337,6 +8551,766 @@
}
]
},
{
"file": "skills/ms-ai-governance/references/responsible-ai/continuous-improvement-feedback-loops.md",
"batch": "R7.4",
"judged_at": "2026-07-31",
"per_file_verdict": "flagged",
"claim_count": 21,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-governance/responsible-ai/continuous-improvement-feedback-loops.md#6",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/en-us/azure/foundry-classic/how-to/monitor-quality-safety?view=foundry-classic",
"evidence_quote": "### Supported metrics for monitoring | Metric | Description | Groundedness ... Relevance ... Coherence ... Fluency",
"reason": "Den kanoniske tabellen som ville listet dem opp har nøyaktig fire monitorerings-scorere (Groundedness, Relevance, Coherence, Fluency) — en femte «Safety»-scorer er fraværende, og Foundrys safety-evaluatorer heter violence/sexual/self_harm/hate_unfairness, ikke «Safety».",
"file": "skills/ms-ai-governance/references/responsible-ai/continuous-improvement-feedback-loops.md",
"line": 74,
"claim": "Microsoft bruker fem automated scorers (LLM-judges) for kontinuerlig kvalitetsvurdering av produksjonstrafikk: Groundedness | Relevance | Coherence | Fluency | Safety.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/continuous-improvement-feedback-loops.md#7",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/foundry-classic/how-to/monitor-quality-safety?view=foundry-classic",
"evidence_quote": "# Set thresholds for passing rate (0.7 = 70%) aggregated_groundedness_pass_rate = 0.7 aggregated_relevance_pass_rate = 0.7 aggregated_coherence_pass_rate = 0.7 aggregated_fluency_pass_rate = 0.7",
"reason": "70 %-delen stemmer nøyaktig for de fire metrikkene, men den bærende «≥ 95 % for safety»-delen finnes ikke — GenerationSafetyQualityMonitoringMetricThreshold konfigurerer kun de fire pass rate-ene, så én bærende del er feil.",
"file": "skills/ms-ai-governance/references/responsible-ai/continuous-improvement-feedback-loops.md",
"line": 78,
"claim": "Eksempel-terskler for scorers: pass rate ≥ 70 % for groundedness, relevance, coherence og fluency, og pass rate ≥ 95 % for safety.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/continuous-improvement-feedback-loops.md#8",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/en-us/azure/databricks/mlflow3/genai/",
"evidence_quote": "MLflow 3 for GenAI provides these key pieces for efficient development, deployment, and continuous improvement: Tracing ... Built-in and custom LLM judges and scorers ... Review apps for expert feedback ... Automated evaluation and monitoring ... App and prompt versioning",
"reason": "Den siterte URL-en (mlflow3/genai/overview/) finnes ikke, og den kanoniske oversiktssiden strukturerer kontinuerlig forbedring som fem komponenter — ingen Learn-side enumererer en «Continuous Improvement Cycle» på 10 navngitte steg.",
"file": "skills/ms-ai-governance/references/responsible-ai/continuous-improvement-feedback-loops.md",
"line": 146,
"claim": "MLflow Continuous Improvement Cycle for GenAI-apper består av 10 steg: Production App | User Feedback | Monitor & Score | Identify Issues | Domain Expert Review | Build Eval Dataset | Tune Scorers | Evaluate New Versions | Compare Results | Deploy or Iterate.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/continuous-improvement-feedback-loops.md#10",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/en-us/azure/foundry-classic/how-to/monitor-quality-safety?view=foundry-classic",
"evidence_quote": "View the metrics on the Token usage tab. ... Select the Generation quality tab to monitor the quality of your application over time. ... To view the operational metrics for the deployment in near real time, select the Operational tab.",
"reason": "De kanoniske sidene enumererer fanene Token usage / Generation quality / Operational (klassisk) og Monitor-fanen med summary cards og «charts and graphs below» (ny Foundry) — verken en «Charts tab» eller en «Logs tab» finnes.",
"file": "skills/ms-ai-governance/references/responsible-ai/continuous-improvement-feedback-loops.md",
"line": 245,
"claim": "Microsoft Foundry monitoring dashboard visualiserer metrics over tid via Charts tab og Logs tab.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/continuous-improvement-feedback-loops.md#19",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/ai-builder/administer-licensing",
"evidence_quote": "AI Builder is licensed on a capacity basis. Building custom and testing models (including prompts) doesn't require AI Builder credits or Copilot Credits. Running them in agents, agent flows, apps, or flows consume these credits.",
"reason": "Inferens-delen stemmer, men den bærende «modelltrening»-delen motsies direkte: trening av modeller krever verken AI Builder credits eller Copilot Credits.",
"file": "skills/ms-ai-governance/references/responsible-ai/continuous-improvement-feedback-loops.md",
"line": 466,
"claim": "AI Builder krever AI Builder credits for modelltrening og inferens.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/continuous-improvement-feedback-loops.md#21",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/microsoft-copilot-studio/guidance/responsible-ai",
"evidence_quote": "",
"reason": "Den siterte responsible-AI-siden omtaler feedback-mekanismer og monitorering, men ingen hentet Learn-side (heller ikke billing-licensing eller analytics-overview) sier at monitoring er inkludert i Copilot Studio-lisensieringen uten separat kostnad.",
"file": "skills/ms-ai-governance/references/responsible-ai/continuous-improvement-feedback-loops.md",
"line": 474,
"claim": "Monitoring-kapabiliteter er inkludert i Microsoft Copilot Studio-lisensieringen, uten separat kostnad for feedback-mekanismer.",
"disposition": "unsourced"
}
]
},
{
"file": "skills/ms-ai-governance/references/responsible-ai/data-quality-responsible-ai.md",
"batch": "R7.4",
"judged_at": "2026-07-31",
"per_file_verdict": "flagged",
"claim_count": 23,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-governance/responsible-ai/data-quality-responsible-ai.md#5",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/well-architected/ai/training-data-design",
"evidence_quote": "Following are some common techniques for preprocessing. This list isn't exhaustive.",
"reason": "Siden dokumenterer fem preprosesseringsteknikker (Quality, Rescoping, Deduplication, Sensitive data handling og Standardized transformation) og sier listen ikke er uttømmende, så det stadfestede antallet fire er feil.",
"file": "skills/ms-ai-governance/references/responsible-ai/data-quality-responsible-ai.md",
"line": 70,
"claim": "Microsoft dokumenterer fire nøkkelteknikker for preprosessering av treningsdata: Quality filtering | Rescoping (broadening overly specific fields) | Deduplication | Sensitive data handling.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/data-quality-responsible-ai.md#9",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/en-us/azure/machine-learning/concept-manage-ml-pitfalls?view=azureml-api-2",
"evidence_quote": "Target leakage is a similar issue. You might not see overfitting between the train and test sets, but the leakage issue appears at prediction-time.",
"reason": "Target leakage er dokumentert som en overfitting-fallgruve, ikke en trussel; verken den siterte sikkerhetssiden eller den kanoniske Threat Modeling AI/ML-siden lister target leakage blant truslene, så trioen finnes ikke i Microsofts trusseloppstilling.",
"file": "skills/ms-ai-governance/references/responsible-ai/data-quality-responsible-ai.md",
"line": 141,
"claim": "Microsoft angir tre trusler mot treningsdata: Malicious data injection | Target leakage | Training data tampering.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/data-quality-responsible-ai.md#11",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/azure/foundry/foundry-models/concepts/models-sold-directly-by-azure#fine-tuning-models",
"evidence_quote": "The following models are supported for fine-tuning:",
"reason": "Den gjeldende fine-tuning-tabellen lister gpt-4o-mini, gpt-4o, gpt-4.1/mini/nano, o4-mini, gpt-5 og enkelte åpne modeller — GPT-4 er ikke med, så påstanden om at GPT-4 kan finjusteres er ikke lenger dekket (kun en utdatert oversiktsside nevner GPT-4 preview).",
"file": "skills/ms-ai-governance/references/responsible-ai/data-quality-responsible-ai.md",
"line": 229,
"claim": "GPT-4 kan finjusteres (fine-tuning) for domenespesifikke oppgaver, for eksempel medisinsk dokumentasjon.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/data-quality-responsible-ai.md#19",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/azure/machine-learning/concept-responsible-ai?view=azureml-api-2",
"evidence_quote": "",
"reason": "Den hentede siden beskriver dashboardets komponenter, men ingen learn.microsoft.com-side sier at Responsible AI Dashboard er inkludert uten ekstra kostnad; prisinformasjon ligger utenfor Learn.",
"file": "skills/ms-ai-governance/references/responsible-ai/data-quality-responsible-ai.md",
"line": 393,
"claim": "Responsible AI Dashboard er inkludert i Azure Machine Learning uten ekstra kostnad.",
"disposition": "unsourced"
},
{
"id": "ms-ai-governance/responsible-ai/data-quality-responsible-ai.md#20",
"judge_verdict": "not_grounded",
"rule": "R3",
"evidence_url": "https://learn.microsoft.com/purview/data-governance-billing",
"evidence_quote": "To use the data governance experience in the Microsoft Purview portal, you need to be set up for Microsoft Purview pay-as-you-go billing.",
"reason": "Data governance i Purview faktureres etter en pay-as-you-go forbruksmodell knyttet til en Azure-abonnement (governed assets per dag og DGPU-er), ikke som en separat lisens eller add-on — rammeverket claimet forutsetter er erstattet.",
"file": "skills/ms-ai-governance/references/responsible-ai/data-quality-responsible-ai.md",
"line": 394,
"claim": "Microsoft Purview krever separat lisens som data governance add-on.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/data-quality-responsible-ai.md#21",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/azure/databricks/admin/account-settings/account",
"evidence_quote": "Azure Databricks is available with two pricing options, Standard and Premium, which offer features for different types of workloads.",
"reason": "Azure Databricks har kun Standard og Premium; det finnes ingen Enterprise-tier, og dokumentasjonen sier at Lakeflow pipelines (som eier expectations) krever Premium-planen, så det oppgitte alternativet Enterprise mangler i den kanoniske oppstillingen.",
"file": "skills/ms-ai-governance/references/responsible-ai/data-quality-responsible-ai.md",
"line": 395,
"claim": "Databricks Expectations krever Databricks-lisens på Premium- eller Enterprise-tier.",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-governance/references/responsible-ai/fairness-testing-measurement.md",
"batch": "R7.4",
"judged_at": "2026-07-31",
"per_file_verdict": "flagged",
"claim_count": 23,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-governance/responsible-ai/fairness-testing-measurement.md#12",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/machine-learning/concept-fairness-ml?view=azureml-api-2",
"evidence_quote": "| ThresholdOptimizer | Postprocessing algorithm based on the paper Equality of Opportunity in Supervised Learning. ... | Binary classification | Categorical | Demographic parity, equalized odds | Post-processing |",
"reason": "Én bærende del er feil: algoritmetabellen oppgir at ThresholdOptimizer støtter demographic parity og equalized odds — equal opportunity er ikke oppført som støttet constraint for noen algoritme, så koblingen «Equal opportunity med ThresholdOptimizer» motsies av kilden selv om de øvrige koblingene stemmer.",
"file": "skills/ms-ai-governance/references/responsible-ai/fairness-testing-measurement.md",
"line": 182,
"claim": "Parity constraints kobles til Fairlearn-algoritmer slik: Demographic parity (binary classification, regression) med ExponentiatedGradient og GridSearch | Equalized odds (binary classification) med ExponentiatedGradient, GridSearch og ThresholdOptimizer | Equal opportunity (binary classification) med ThresholdOptimizer | Bounded group loss (regression) med GridSearch.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/fairness-testing-measurement.md#17",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/python/api/azure-ai-evaluation/azure.ai.evaluation.hateunfairnessevaluator?view=azure-python",
"evidence_quote": "Note: This is an experimental class, and may change at any time. Please see https://aka.ms/azuremlexperimental for more information.",
"reason": "Rollen som content safety-evaluator stemmer, men den bærende status-delen er feil: API-siden merker klassen som eksperimentell, ikke GA.",
"file": "skills/ms-ai-governance/references/responsible-ai/fairness-testing-measurement.md",
"line": 336,
"claim": "HateUnfairnessEvaluator er GA i Microsoft Foundry som content safety evaluator for generative modeller.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/fairness-testing-measurement.md#19",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/ai-builder/overview",
"evidence_quote": "",
"reason": "Verken den siterte siden eller søk i AI Builder-dokumentasjonen (overview, model-types, administer, responsible AI FAQ) omtaler fairness assessment-UI for custom models eller at Microsoft utfører fairness-testing av prebuilt-modeller — påstanden kan verken bekreftes eller avkreftes.",
"file": "skills/ms-ai-governance/references/responsible-ai/fairness-testing-measurement.md",
"line": 350,
"claim": "Power Platform AI Builder har ingen innebygd fairness assessment-UI for custom models (per 2026-02), mens fairness testing for pre-built models er utført av Microsoft.",
"disposition": "unsourced"
},
{
"id": "ms-ai-governance/responsible-ai/fairness-testing-measurement.md#20",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/azure/machine-learning/concept-fairness-ml?view=azureml-api-2",
"evidence_quote": "",
"reason": "Learn omtaler Fairlearn konsekvent som «the Fairlearn open-source package», men ingen learn.microsoft.com-side oppgir lisensen MIT — den konkrete verdien i claimet er ikke dokumentert.",
"file": "skills/ms-ai-governance/references/responsible-ai/fairness-testing-measurement.md",
"line": 399,
"claim": "Fairlearn er gratis open-source under MIT License.",
"disposition": "unsourced"
},
{
"id": "ms-ai-governance/responsible-ai/fairness-testing-measurement.md#22",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/azure/databricks/data-governance/unity-catalog/data-quality-monitoring/data-profiling/",
"evidence_quote": "## Requirements - Your workspace must be enabled for Unity Catalog and you must have access to Databricks SQL. - To enable data profiling, you must have the following privileges: USE CATALOG on the catalog and USE SCHEMA on the schema containing the table. SELECT on the table. MANAGE on the catalog, schema, or table.",
"reason": "Den kanoniske Requirements-listen for data profiling (som fairness- og bias-metrikkene hører til) nevner ingen Premium- eller Enterprise-tier — kravet er Unity Catalog og Databricks SQL; i tillegg er rammen «Lakehouse Monitoring» erstattet («Data profiling was formerly known as Lakehouse Monitoring»).",
"file": "skills/ms-ai-governance/references/responsible-ai/fairness-testing-measurement.md",
"line": 424,
"claim": "Fairness metrics i Databricks er del av Lakehouse monitoring og krever Databricks Premium- eller Enterprise-tier.",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-governance/references/responsible-ai/human-in-the-loop-oversight.md",
"batch": "R7.4",
"judged_at": "2026-07-31",
"per_file_verdict": "flagged",
"claim_count": 24,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-governance/responsible-ai/human-in-the-loop-oversight.md#1",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/microsoft-copilot-studio/flows-advanced-approvals",
"evidence_quote": "In your flow, you can add *Run a multistage approval* as an action via the new *Human review* connector.",
"reason": "Preview-statusen og conditions mellom stages stemmer, men den nye koblingen heter *Human review*, ikke 'Human in the loop' — en lastbærende, navngitt streng leseren ville søkt etter i handlingslisten.",
"file": "skills/ms-ai-governance/references/responsible-ai/human-in-the-loop-oversight.md",
"line": 48,
"claim": "Multistage- og AI-approvals i Power Automate/Copilot Studio er i Preview, og det finnes en ny 'Human in the loop'-kobling samt conditions mellom stages for dynamisk routing.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/human-in-the-loop-oversight.md#2",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/microsoft-copilot-studio/faqs-ai-approvals",
"evidence_quote": "AI approvals support GPT 4.1 mini, GPT 4.o, GPT 4.1, and o3 models, which are hosted on Azure OpenAI Service.",
"reason": "Approve/Reject med begrunnelse stemmer, men modellen er ikke låst til GPT-o3: fire modeller støttes via en model picker, og FAQ-en sier GPT-4.1 er typisk ideell — den lastbærende versjonsdelen holder ikke.",
"file": "skills/ms-ai-governance/references/responsible-ai/human-in-the-loop-oversight.md",
"line": 48,
"claim": "AI-stage i Power Automate-approvals bruker GPT-o3 til å gjøre Approve/Reject med begrunnelse.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/human-in-the-loop-oversight.md#5",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/en-us/power-automate/all-assigned-must-approve",
"evidence_quote": "Approve/Reject - Everyone must approve | Approve/Reject - First to respond | Custom responses - Wait for all responses | Custom responses - Wait for one response",
"reason": "Den kanoniske tabellen 'Approval types and their behaviors' lister fire typer, og verken 'Conditional approvals' eller 'Multistage' står der — de er egenskaper ved multistage-flyten, ikke godkjenningstyper; de to Custom responses-typene mangler i claimet.",
"file": "skills/ms-ai-governance/references/responsible-ai/human-in-the-loop-oversight.md",
"line": 55,
"claim": "Godkjenningstypene i approval-flyter er: First to respond | Everyone must approve | Conditional approvals | Multistage.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/human-in-the-loop-oversight.md#7",
"judge_verdict": "not_grounded",
"rule": "R7",
"evidence_url": "https://learn.microsoft.com/agent-framework/integrations/ag-ui/human-in-the-loop",
"evidence_quote": "1. always_require: Always request approval before execution 2. never_require: Never request approval (default behavior) 3. conditional: Request approval based on certain conditions (custom logic)",
"reason": "Enum-verdiene er lastbærende strenger: default-modusen heter never_require (ikke 'never'), og den tredje modusen heter conditional — 'confidence_based' finnes ikke i API-et (Literal['always_require', 'never_require']).",
"file": "skills/ms-ai-governance/references/responsible-ai/human-in-the-loop-oversight.md",
"line": 90,
"claim": "Microsoft Agent Framework har approval modes for funksjoner: `never` (default) | `always_require` | `confidence_based`.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/human-in-the-loop-oversight.md#8",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/en-us/security/benchmark/azure/mcsb-v2-artificial-intelligence-security#ai-5-ensure-human-in-the-loop",
"evidence_quote": "Secure HITL interfaces: Protect review systems with encryption, implement strict access controls using Microsoft Entra ID, and deploy anomaly detection to prevent tampering or unauthorized access to approval processes.",
"reason": "AI-5.1s sikkerhetskrav sier 'encryption' generelt; ingen TLS-versjon er angitt noe sted i kontrollen — kravet om 'TLS 1.2 eller nyere' er lagt til av claimet.",
"file": "skills/ms-ai-governance/references/responsible-ai/human-in-the-loop-oversight.md",
"line": 128,
"claim": "Sikkerhetskravet i AI-5.1 krever kryptering av review-systemer med TLS 1.2 eller nyere.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/human-in-the-loop-oversight.md#9",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/en-us/security/benchmark/azure/mcsb-v2-artificial-intelligence-security#ai-5-ensure-human-in-the-loop",
"evidence_quote": "Define critical actions: Identify high-risk AI operations requiring human review such as external data transfers, processing of confidential information, or decisions impacting financial or operational outcomes, using risk assessments to prioritize review pathways.",
"reason": "AI-5 definerer ikke fem kriterier som 'alltid krever HITL', men gir tre eksempler innledet med 'such as'; 'Safety-related commands' og 'Compliance-critical processes' står ikke i oppregningen, og 'operational outcomes' er utelatt.",
"file": "skills/ms-ai-governance/references/responsible-ai/human-in-the-loop-oversight.md",
"line": 310,
"claim": "Azure AI Security Benchmark AI-5 definerer fem kriterier som alltid krever HITL: External data transfers | Processing of confidential information | Decisions impacting financial outcomes | Safety-related commands | Compliance-critical processes.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/human-in-the-loop-oversight.md#10",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/en-us/security/benchmark/azure/mcsb-v2-artificial-intelligence-security#ai-5-ensure-human-in-the-loop",
"evidence_quote": "Train reviewers: Equip personnel with training on AI system behavior, potential vulnerabilities (e.g., adversarial inputs), and domain-specific risks, providing access to contextual data and decision-support tools to enable informed validation.",
"reason": "Opplæringskravet omfatter tre temaer pluss tilgang til kontekstdata og beslutningsstøtteverktøy — 'Escalation procedures' finnes ikke i AI-5.1, så den femte oppføringen er uten dekning.",
"file": "skills/ms-ai-governance/references/responsible-ai/human-in-the-loop-oversight.md",
"line": 349,
"claim": "AI-5.1 sine opplæringskrav for reviewere omfatter fem punkter: AI system behavior | Potential vulnerabilities | Domain-specific risks | Decision-support tools | Escalation procedures.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/human-in-the-loop-oversight.md#11",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/en-us/security/benchmark/azure/mcsb-v2-artificial-intelligence-security#ai-5-ensure-human-in-the-loop",
"evidence_quote": "Optimize review processes: Implement selective HITL reviewing only low-confidence AI outputs or high-impact decisions to balance security with operational efficiency, regularly assessing workflows to prevent reviewer fatigue and maintain effectiveness.",
"reason": "Kontrollen som faktisk adresserer reviewer fatigue foreskriver selektiv review og jevnlig vurdering av arbeidsflyten, ikke noe tak; grensen på 50 beslutninger per dag står ingen steder.",
"file": "skills/ms-ai-governance/references/responsible-ai/human-in-the-loop-oversight.md",
"line": 356,
"claim": "For å forhindre reviewer fatigue skal en person ikke review mer enn 50 AI-beslutninger per dag.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/human-in-the-loop-oversight.md#13",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/en-us/microsoft-copilot-studio/guidance/deflection-topic-escalation-analysis",
"evidence_quote": "The direct way to initiate an escalation to a human representative is through the **Escalate** system topic. ... Another way to trigger this escalation is through the **Transfer Conversation** node in the authoring canvas.",
"reason": "Escalate-systemtopic mot Omnichannel og 'Escalation Rate Drivers' i analytics stemmer, men den kanoniske gjennomgangen kjenner verken 'Approval Topics' eller 'Rationale Generation' — godkjenninger i Copilot Studio er stages i agent flows via Human review-koblingen, ikke topics.",
"file": "skills/ms-ai-governance/references/responsible-ai/human-in-the-loop-oversight.md",
"line": 396,
"claim": "Copilot Studio sine HITL-features er: Human Handoff Topic som overfører samtalen til Live Agent (Omnichannel, Dynamics 365) | Escalation Rate Tracking i analytics dashboard | Rationale Generation | Approval Topics som pauser for menneskelig input.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/human-in-the-loop-oversight.md#15",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/microsoft-copilot-studio/faqs-ai-approvals",
"evidence_quote": "GPT-4.1 is typically ideal for most approval scenarios. Advanced reasoning models (like O3) might handle complex logic better but are slower.",
"reason": "O3-delen stemmer, men FAQ-en anbefaler GPT-4.1 for de fleste scenarioene og tilskriver nettopp kompleks logikk til O3 — claimet snur dette og gir GPT-4.1 de komplekse approval-scenarioene.",
"file": "skills/ms-ai-governance/references/responsible-ai/human-in-the-loop-oversight.md",
"line": 421,
"claim": "FAQ for AI Approvals anbefaler GPT-4.1 for komplekse approval-scenarioer, og o3 for advanced reasoning men da tregere.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/human-in-the-loop-oversight.md#19",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/security/benchmark/azure/mcsb-v2-artificial-intelligence-security#ai-5-ensure-human-in-the-loop",
"evidence_quote": "",
"reason": "Den siterte MCSB-siden sier ingenting om tilgjengelighet, og ingen learn.microsoft.com-side slår fast at Power Automate Approvals er skjermleser-kompatibel, at Azure Dashboards er WCAG 2.1 AA-compliant eller at Copilot Studio har keyboard navigation support; slik samsvarsinformasjon publiseres i ACR-rapporter utenfor Learn.",
"file": "skills/ms-ai-governance/references/responsible-ai/human-in-the-loop-oversight.md",
"line": 585,
"claim": "Microsofts tilgjengelighets-features for HITL-grensesnitt: Power Automate Approvals er skjermleser-kompatibel | Azure Dashboards er WCAG 2.1 AA-compliant | Copilot Studio har keyboard navigation support.",
"disposition": "unsourced"
},
{
"id": "ms-ai-governance/responsible-ai/human-in-the-loop-oversight.md#21",
"judge_verdict": "not_grounded",
"rule": "R3",
"evidence_url": "https://learn.microsoft.com/ai-builder/administer-licensing",
"evidence_quote": "Agents and agent flows only consume Copilot Credits.",
"reason": "Valutaen ble byttet fra messages til Copilot Credits 1. september 2025, og AI Builder credits er en egen valuta som agent flows nettopp ikke bruker — enheten claimet bygger på er erstattet.",
"file": "skills/ms-ai-governance/references/responsible-ai/human-in-the-loop-oversight.md",
"line": 604,
"claim": "AI Approvals er inkludert i Copilot Studio-lisensen (per-user-lisens + AI credits) og forbruker AI credits ved bruk.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/human-in-the-loop-oversight.md#22",
"judge_verdict": "not_grounded",
"rule": "",
"evidence_url": "https://learn.microsoft.com/power-automate/get-started-approvals",
"evidence_quote": "Because the approvals connector is a standard connector, any license that grants access to Power Automate and the ability to use standard connectors is sufficient to create approval flows.",
"reason": "Kilden sier det motsatte: approvals er en standardkobling, og Office 365- eller Dynamics 365-lisenser holder — Power Automate Premium er ikke et krav for approvals.",
"file": "skills/ms-ai-governance/references/responsible-ai/human-in-the-loop-oversight.md",
"line": 628,
"claim": "Power Automate Premium kreves for approvals.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/human-in-the-loop-oversight.md#23",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/agent-framework/overview/",
"evidence_quote": "Supports Microsoft Foundry, Anthropic, Azure OpenAI, OpenAI, Ollama, and more.",
"reason": "Open source-delen står seg, men den lastbærende påstanden om at Azure OpenAI eller Foundry kreves motsies: rammeverket støtter også OpenAI direkte, Anthropic og lokal Ollama-kjøring.",
"file": "skills/ms-ai-governance/references/responsible-ai/human-in-the-loop-oversight.md",
"line": 637,
"claim": "Microsoft Agent Framework har ingen direkte lisenskostnad (open source), men krever Azure OpenAI eller Microsoft Foundry for modellene.",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-governance/references/responsible-ai/model-explainability-interpretability.md",
"batch": "R7.4",
"judged_at": "2026-07-31",
"per_file_verdict": "flagged",
"claim_count": 23,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-governance/responsible-ai/model-explainability-interpretability.md#1",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/azure/machine-learning/how-to-responsible-ai-insights-ui?view=azureml-api-2",
"evidence_quote": "This feature is currently in public preview. This preview version is provided without a service-level agreement, and we don't recommend it for production workloads.",
"reason": "Blankt GA-stempel holder ikke: cloud-parity-tabellen gir Interpretability SDK status GA, men Explainability in UI og Fairness-integrasjonen står som Preview, og selve genereringen av Responsible AI dashboard/scorecard i studio er eksplisitt merket public preview.",
"file": "skills/ms-ai-governance/references/responsible-ai/model-explainability-interpretability.md",
"line": 4,
"claim": "Azure Machine Learning sine model interpretability-/Responsible AI-kapabiliteter har status GA (generelt tilgjengelig).",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/model-explainability-interpretability.md#11",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/fabric/data-science/explainable-boosting-machines-classification",
"evidence_quote": "",
"reason": "Fabric-siden beskriver at SynapseML-EBM forventer input som en vektor av Doubles og bruker StringIndexer for kategoriske variabler, men ingen learn.microsoft.com-side sier at EBM er begrenset til tabulære data — begrensningen er verken bekreftet eller motsagt.",
"file": "skills/ms-ai-governance/references/responsible-ai/model-explainability-interpretability.md",
"line": 194,
"claim": "Explainable Boosting Machines (EBM) fra InterpretML er begrenset til tabulære data.",
"disposition": "unsourced"
},
{
"id": "ms-ai-governance/responsible-ai/model-explainability-interpretability.md#13",
"judge_verdict": "not_grounded",
"rule": "",
"evidence_url": "https://learn.microsoft.com/azure/machine-learning/how-to-use-serverless-compute?view=azureml-api-2",
"evidence_quote": "You can also use serverless compute to build environment images and for responsible AI dashboard scenarios.",
"reason": "Kilden sier det motsatte av påstanden: serverless compute støttes nettopp for Responsible AI dashboard-scenarier, og AutoML-veiledningen ber deg velge Serverless som compute når du genererer dashboardet — compute cluster er altså ikke et krav.",
"file": "skills/ms-ai-governance/references/responsible-ai/model-explainability-interpretability.md",
"line": 265,
"claim": "Generering av forklaringer i Responsible AI dashboard krever compute cluster, ikke serverless compute.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/model-explainability-interpretability.md#18",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/ai-builder/prediction-performance",
"evidence_quote": "After each training, a list of top influencers appears on the model details page. Each column used in the training has a score to represent its influence on the training. These scores combine to equal 100 percent.",
"reason": "Første del stemmer (AI Builder gir confidence scores), men den andre bærende delen er motsagt: AI Builder prediction gir nettopp forklaringer på feature-nivå gjennom top influencers med score per kolonne.",
"file": "skills/ms-ai-governance/references/responsible-ai/model-explainability-interpretability.md",
"line": 288,
"claim": "Power Platform AI Builder gir prediction confidence scores, men ikke feature-level explanations.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/model-explainability-interpretability.md#19",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/azure/machine-learning/concept-responsible-ai",
"evidence_quote": "",
"reason": "Den siterte siden sier ingenting om regioner, og ingen learn.microsoft.com-side lister Azure Machine Learning-tilgjengelighet for Norway East og Norway West — regional tilgjengelighet ligger på den JS-genererte Products available by region-siden utenfor Learn.",
"file": "skills/ms-ai-governance/references/responsible-ai/model-explainability-interpretability.md",
"line": 347,
"claim": "Azure ML kan kjøre SHAP-beregninger i Norge-regionene Norway East og Norway West.",
"disposition": "unsourced"
},
{
"id": "ms-ai-governance/responsible-ai/model-explainability-interpretability.md#22",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/azure/machine-learning/concept-responsible-ai",
"evidence_quote": "",
"reason": "Ingen learn.microsoft.com-side stiller opp denne fire-veis lisensmodell-matrisen; prisingsmodellene (pay-as-you-go, pay-per-token, per user/per capacity) ligger på JS-genererte Azure-prissider, så påstanden kan verken bekreftes eller motbevises herfra.",
"file": "skills/ms-ai-governance/references/responsible-ai/model-explainability-interpretability.md",
"line": 390,
"claim": "Lisensmodellene for XAI-funksjonalitet er: Azure Machine Learning = pay-as-you-go (compute + storage, gir Responsible AI Dashboard, InterpretML, SHAP) | Microsoft Fabric = capacity-based CU per måned (gir TabularSHAP i Synapse ML og EBM i InterpretML) | Power BI Premium = per user eller per capacity (kan visualisere SHAP-data fra Fabric/AML) | Microsoft Foundry = pay-per-token (gir GPT-assisterte evaluators).",
"disposition": "unsourced"
},
{
"id": "ms-ai-governance/responsible-ai/model-explainability-interpretability.md#23",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/azure/cost-management-billing/manage/create-free-services",
"evidence_quote": "",
"reason": "Learn-siden om tjenester inkludert i Azure free account nevner ingen Azure Machine Learning-kvote på 4 timer compute per måned og henviser videre til den JS-genererte Azure free account FAQ-en; verdien er verken bekreftet eller motsagt på Learn.",
"file": "skills/ms-ai-governance/references/responsible-ai/model-explainability-interpretability.md",
"line": 396,
"claim": "Azure ML free tier gir 4 timer compute per måned og kan brukes til utvikling av XAI.",
"disposition": "unsourced"
}
]
},
{
"file": "skills/ms-ai-governance/references/responsible-ai/model-monitoring-drift-detection.md",
"batch": "R7.4",
"judged_at": "2026-07-31",
"per_file_verdict": "flagged",
"claim_count": 30,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-governance/responsible-ai/model-monitoring-drift-detection.md#2",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/en-us/azure/machine-learning/concept-model-monitoring?view=azureml-api-2",
"evidence_quote": "Built-in monitoring signals for tabular data, including data drift, prediction drift, data quality, feature attribution drift, and model performance.",
"reason": "Den kanoniske opplistingen av signaler i Azure ML model monitoring inneholder ikke «concept drift»; concept drift står kun i WAF-artikkelen som et generelt model decay-begrep, så den påståtte fem-delingen finnes ikke i produktet.",
"file": "skills/ms-ai-governance/references/responsible-ai/model-monitoring-drift-detection.md",
"line": 33,
"claim": "Azure ML model monitoring skiller mellom fem kjernetyper av drift: Data drift | Concept drift | Prediction drift | Data quality drift | Feature attribution drift.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/model-monitoring-drift-detection.md#3",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/machine-learning/concept-model-monitoring?view=azureml-api-2",
"evidence_quote": "Built-in monitoring signals for tabular data, including data drift, prediction drift, data quality, feature attribution drift, and model performance.",
"reason": "Siden oppgir fem innebygde signaler for tabular data, og plasserer Generation safety/quality under «Questions & Answers» (generativ AI) — ikke tabulært — så den løftbærende delen «seks … for tabular data» er motsagt.",
"file": "skills/ms-ai-governance/references/responsible-ai/model-monitoring-drift-detection.md",
"line": 47,
"claim": "Azure Machine Learning tilbyr seks innebygde monitoring-signaler for tabular data: Data drift | Prediction drift | Data quality | Feature attribution drift | Model performance | Generation safety/quality.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/model-monitoring-drift-detection.md#10",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/machine-learning/concept-model-monitoring?view=azureml-api-2",
"evidence_quote": "Groundedness, Relevance, Fluency, Similarity, Coherence | Questions & Answers | Prompt, completion, context, and annotation template | N/A",
"reason": "Metrikkene og Q&A-oppgaven stemmer, men annotation template står i produksjonsdata-kolonnen og referansedata er oppgitt som N/A — den løftbærende påstanden «bruker annotation template som referansedata» er motsagt.",
"file": "skills/ms-ai-governance/references/responsible-ai/model-monitoring-drift-detection.md",
"line": 56,
"claim": "Generation safety/quality-signalet bruker metrikkene Groundedness | Relevance | Fluency | Similarity | Coherence, gjelder generativ AI (Q&A) og bruker annotation template som referansedata.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/model-monitoring-drift-detection.md#14",
"judge_verdict": "not_grounded",
"rule": "R4",
"evidence_url": "https://learn.microsoft.com/en-us/azure/machine-learning/how-to-monitor-model-performance?view=azureml-api-2",
"evidence_quote": "instance_type: standard_e4s_v3\n runtime_version: \"3.4\"",
"reason": "Alle konfigurasjonseksemplene på den gjeldende v2-siden (CLI og SDK) setter runtime_version 3.4; 3.3 står kun i den utdaterte v1-artikkelen, og Synapse Spark 3.3 hadde end of support 31. mars 2025.",
"file": "skills/ms-ai-governance/references/responsible-ai/model-monitoring-drift-detection.md",
"line": 103,
"claim": "Serverless Spark compute for Azure ML model monitoring konfigureres med runtime_version 3.3.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/model-monitoring-drift-detection.md#20",
"judge_verdict": "not_grounded",
"rule": "R4",
"evidence_url": "https://learn.microsoft.com/en-us/azure/machine-learning/how-to-monitor-datasets?view=azureml-api-1",
"evidence_quote": "Data drift (preview) was retired on September 1, 2025. Migrate to Model Monitor for your data drift tasks.",
"reason": "Beskrivelsen (metrics i json til blob storage, Application Insights for custom alerting på alle metrics) står kun på den pensjonerte v1-datasettmonitorsiden; den gjeldende v2-dokumentasjonen nevner ikke Application Insights for model monitoring, og Azure Monitor Metrics-fanen brukes der til å se Event Grid-events, ikke performance.",
"file": "skills/ms-ai-governance/references/responsible-ai/model-monitoring-drift-detection.md",
"line": 411,
"claim": "Monitoring metrics sendes til Azure Blob Storage i JSON-format, med Application Insights for custom alerting på alle metrics og Azure Monitor Metrics for performance-visualisering.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/model-monitoring-drift-detection.md#25",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/virtual-machines/sizes/memory-optimized/ev3-esv3-series",
"evidence_quote": "| Standard_E64s_v3 | 64 | 432 | 864 | 32 | 128000/1024 (1600) | 80000/1200 | 80000/2000 | 8/30000 |",
"reason": "De fire første størrelsene stemmer (4/32, 8/64, 16/128, 32/256), men Standard_E64s_v3 har 432 GiB minne, ikke 512 GB — én løftbærende del er feil.",
"file": "skills/ms-ai-governance/references/responsible-ai/model-monitoring-drift-detection.md",
"line": 513,
"claim": "Serverless Spark-størrelsene har spesifikasjonene standard_e4s_v3 4 vCPU/32 GB | standard_e8s_v3 8 vCPU/64 GB | standard_e16s_v3 16 vCPU/128 GB | standard_e32s_v3 32 vCPU/256 GB | standard_e64s_v3 64 vCPU/512 GB.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/model-monitoring-drift-detection.md#26",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/azure/machine-learning/concept-plan-manage-cost?view=azureml-api-2",
"evidence_quote": "",
"reason": "Verken konseptsiden for model monitoring eller kostnadssiden sier at model monitoring er uten ekstra lisens/kostnad utover compute; prisingen ligger på azure.microsoft.com-prissiden som ikke er en Learn-kilde.",
"file": "skills/ms-ai-governance/references/responsible-ai/model-monitoring-drift-detection.md",
"line": 550,
"claim": "Model monitoring er inkludert i Azure ML uten ekstra lisens eller kostnad utover compute; Azure ML workspace krever kun Azure subscription.",
"disposition": "unsourced"
},
{
"id": "ms-ai-governance/responsible-ai/model-monitoring-drift-detection.md#27",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/azure/machine-learning/concept-plan-manage-cost?view=azureml-api-2",
"evidence_quote": "",
"reason": "Ingen Learn-side oppgir faktureringsmodellen for Azure ML serverless Spark som pay-per-use per vCPU-time; kostnadssiden dekker kun VM-/compute-instance-fakturering per time.",
"file": "skills/ms-ai-governance/references/responsible-ai/model-monitoring-drift-detection.md",
"line": 551,
"claim": "Serverless Spark for monitoring faktureres pay-per-use og belastes per vCPU-time.",
"disposition": "unsourced"
}
]
},
{
"file": "skills/ms-ai-governance/references/responsible-ai/red-teaming-ai-models.md",
"batch": "R7.4",
"judged_at": "2026-07-31",
"per_file_verdict": "flagged",
"claim_count": 24,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-governance/responsible-ai/red-teaming-ai-models.md#2",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/azure/foundry/openai/concepts/red-teaming",
"evidence_quote": "",
"reason": "Den siterte siden nevner ikke PyRIT i det hele tatt, og søk på learn.microsoft.com finner ingen side som lister PyRIT-komponentene Prompt Executors/Datasets/Converters/Scorers/Memory/Targets — komponentdokumentasjonen ligger utenfor Learn (github.io/PyRIT), så påstanden kan verken bekreftes eller motbevises der.",
"file": "skills/ms-ai-governance/references/responsible-ai/red-teaming-ai-models.md",
"line": 48,
"claim": "PyRIT (Python Risk Identification Tool for generative AI) består av komponentene Prompt Executors | Datasets | Converters | Scorers | Memory | Targets.",
"disposition": "unsourced"
},
{
"id": "ms-ai-governance/responsible-ai/red-teaming-ai-models.md#3",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/azure/foundry/how-to/develop/run-scans-ai-red-teaming-agent",
"evidence_quote": "",
"reason": "Learn viser kun at RedTeam kan scanne tekstbaserte PyRIT PromptChatTarget og lenker videre til PyRITs egen target-liste utenfor Learn; ingen Learn-side lister integrasjonene Azure OpenAI/Hugging Face/REST APIs/lokale modeller for PyRIT Targets.",
"file": "skills/ms-ai-governance/references/responsible-ai/red-teaming-ai-models.md",
"line": 53,
"claim": "PyRIT Targets støtter integrasjoner mot Azure OpenAI | Hugging Face | REST APIs | lokale modeller.",
"disposition": "unsourced"
},
{
"id": "ms-ai-governance/responsible-ai/red-teaming-ai-models.md#11",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/foundry/concepts/evaluation-evaluators/risk-safety-evaluators",
"evidence_quote": "| Indirect Attack (XPIA) | Model only | Measures to what extent the response fell for the indirect jailbreak attempt.",
"reason": "Prohibited actions, sensitive data leakage og task adherence er korrekt merket «Agents only»/«Cloud only», men den fjerde bærende delen svikter: XPIA er dokumentert som «Model only» i Risk and Safety Evaluations og IndirectAttack (XPIA) er dessuten tilgjengelig som angrepsstrategi i lokale scans — altså verken kun for agenter eller kun i sky.",
"file": "skills/ms-ai-governance/references/responsible-ai/red-teaming-ai-models.md",
"line": 98,
"claim": "Risikokategoriene Prohibited Actions | Sensitive Data Leakage | Task Adherence | Indirect Prompt Injection (XPIA) støttes kun for agenter og kun i cloud red teaming.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/red-teaming-ai-models.md#17",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/security/benchmark/azure/mcsb-v2-artificial-intelligence-security",
"evidence_quote": "",
"reason": "MCSB AI-7 anbefaler både PyRIT og AI Red Teaming Agent, CI/CD-integrasjon, Azure Monitor og månedlig/kvartalsvis kadens, og agentsiden lister fasene design/development/pre-deployment/post-deployment — men ingen Learn-side kobler verktøy (og særlig sky- vs. lokal-modus) til enkeltfaser slik påstandens tabell gjør.",
"file": "skills/ms-ai-governance/references/responsible-ai/red-teaming-ai-models.md",
"line": 156,
"claim": "Microsofts anbefalte red teaming-verktøy per fase: Design → AI Red Teaming Agent (cloud) | Development → PyRIT (local) med CI/CD-integrasjon | Pre-deployment → AI Red Teaming Agent (cloud) | Post-deployment → AI Red Teaming Agent (cloud) + Azure Monitor med månedlig/kvartalsvis frekvens.",
"disposition": "unsourced"
},
{
"id": "ms-ai-governance/responsible-ai/red-teaming-ai-models.md#22",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/azure/foundry/openai/concepts/red-teaming",
"evidence_quote": "",
"reason": "Learn omtaler PyRIT konsekvent som «Microsoft's open-source framework» og lenker til GitHub, men ingen learn.microsoft.com-side oppgir lisenstypen MIT eller at bruken er gratis — lisensvilkår er ikke noe Learn ville regnes å oppgi for et GitHub-prosjekt.",
"file": "skills/ms-ai-governance/references/responsible-ai/red-teaming-ai-models.md",
"line": 401,
"claim": "PyRIT-rammeverket er lisensiert under MIT License og er gratis å bruke.",
"disposition": "unsourced"
},
{
"id": "ms-ai-governance/responsible-ai/red-teaming-ai-models.md#23",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/foundry/responsible-ai/openai/limited-access",
"evidence_quote": "Unless otherwise indicated in the service, all Azure customers are eligible for access to Models sold by Azure ... so customers are not required to submit a registration form unless they are: (a) accessing a model sold by Azure designated as a Limited Access Service, or (b) requesting approval to modify Guardrails",
"reason": "Den bærende delen «Azure OpenAI Service krever Azure subscription + godkjent søknad» er motsagt: registreringsskjema kreves ikke lenger generelt, kun for modeller som er utpekt som Limited Access eller for modifiserte guardrails/abuse monitoring.",
"file": "skills/ms-ai-governance/references/responsible-ai/red-teaming-ai-models.md",
"line": 424,
"claim": "Minimum lisenskrav: Microsoft Foundry krever Azure subscription (Pay-As-You-Go eller Enterprise Agreement) | Azure OpenAI Service krever Azure subscription + godkjent søknad | Azure AI Content Safety er inkludert i Azure AI Services (pay-per-transaction) | PyRIT krever ingen lisens (MIT open-source).",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-governance/references/responsible-ai/responsible-ai-framework-overview.md",
"batch": "R7.4",
"judged_at": "2026-07-31",
"per_file_verdict": "flagged",
"claim_count": 24,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-governance/responsible-ai/responsible-ai-framework-overview.md#16",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/en-us/power-platform/responsible-ai-overview",
"evidence_quote": "This app contains a growing list of AI-driven features. ... Copilot data security and privacy, Copilot data sharing, Use Copilot to generate deployment notes in pipelines",
"reason": "Den kanoniske RAI-oversikten for Power Platform lister ingen fairness-vurdering for AI Builder-modeller og ingen RAI-vurdering av Power Automate-workflows; Power Platform admin center dokumenterer kredittforbruk/kapasitet, ikke drift-tracking — de tre navngitte kapabilitetene finnes ikke.",
"file": "skills/ms-ai-governance/references/responsible-ai/responsible-ai-framework-overview.md",
"line": 158,
"claim": "Power Platform AI dekker RAI gjennom AI Builder (fairness-vurderinger for AI Builder-modeller) | Power Automate (RAI-vurderinger for workflows med AI-komponenter) | drift-/performance-tracking via Power Platform Admin Center.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/responsible-ai-framework-overview.md#18",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/azure/machine-learning/concept-responsible-ai-dashboard",
"evidence_quote": "",
"reason": "Ingen Learn-side sier at dashbordet er lisensfritt eller inkludert i workspace-kostnadene (pris/lisens er ikke opplyst); compute-instance-delen bekreftes riktignok av how-to-sidens forutsetninger, men selve kostnadspåstanden kan verken bekreftes eller avkreftes.",
"file": "skills/ms-ai-governance/references/responsible-ai/responsible-ai-framework-overview.md",
"line": 209,
"claim": "Azure Machine Learning Responsible AI Dashboard krever ingen separat lisens — den er inkludert i Azure ML workspace-kostnadene, og komponentene kjører på Azure ML compute instances (CPU/GPU).",
"disposition": "unsourced"
},
{
"id": "ms-ai-governance/responsible-ai/responsible-ai-framework-overview.md#19",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/compliance/assurance/assurance-artificial-intelligence",
"evidence_quote": "Copilot interactions are stored within the user's Exchange Online mailbox. This data includes: Prompt and response pairs, Metadata for auditing and compliance, Logs for Copilot Safety Data (CSD) analysis",
"reason": "«Zero data retention» er en bærende del som motsies direkte — interaksjoner lagres i mailboksen med konfigurerbar retention; i tillegg lister lisenskravssiden langt flere kvalifiserende baselisenser enn bare E3/E5 (A3/A5, Business-planer, F1/F3, Office 365 E1/E3/E5).",
"file": "skills/ms-ai-governance/references/responsible-ai/responsible-ai-framework-overview.md",
"line": 220,
"claim": "Microsoft 365 Copilot krever Microsoft 365 E3 eller E5 pluss en Copilot-lisens; RAI-funksjonene (grounding, safety filters, zero data retention, Purview governance) er inkludert.",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/responsible-ai-framework-overview.md#23",
"judge_verdict": "not_grounded",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/azure/ai-foundry/responsible-ai/openai/data-privacy",
"evidence_quote": "If the customer has been approved for modified abuse monitoring (learn more at Abuse Monitoring), the data storage and human review process described above is not performed.",
"reason": "Standardoppsettet er det motsatte av zero data retention: prompts og completions lagres i abuse monitoring-datalageret med mindre kunden er godkjent for modified abuse monitoring, som må søkes om.",
"file": "skills/ms-ai-governance/references/responsible-ai/responsible-ai-framework-overview.md",
"line": 260,
"claim": "Azure OpenAI har zero data retention som standardinnstilling (bør verifiseres i Product Terms).",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-security/references/ai-security-engineering/ai-prompt-shield-network.md",
"batch": "R7.1",