open/ms-ai-architect
1
0
Fork 0
Code Issues Pull requests Releases Wiki Activity Actions
ms-ai-architect/scripts/kb-eval/data/judge-pass-manifest.json

1613 lines
124 KiB
JSON
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

{
"_meta": {
"purpose": "R7–R10 full-pass judge ledger: one record per never-verified reference file — born-verified pass (surgical stamp) or flagged (R11 work-list). Durable per file.",
"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": 30,
"generated": "2026-07-18"
},
"files": [
{
"file": "skills/ms-ai-engineering/references/agent-orchestration/foundry-workflows-visual-orchestration.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 16,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-engineering/agent-orchestration/foundry-workflows-visual-orchestration.md#2",
"judge_verdict": "not_grounded",
"rule": "R3",
"evidence_url": "https://learn.microsoft.com/azure/foundry/agents/concepts/workflow",
"evidence_quote": "Because workflows are no longer a separate agent type, you run the result by deploying it as a hosted agent",
"reason": "The claim's three-agent-type taxonomy (Prompt-based | Workflow | Hosted) is superseded: the live docs state workflows are no longer a separate agent type and the development-lifecycle page enumerates exactly two agent types (Prompt-based and Hosted).",
"file": "skills/ms-ai-engineering/references/agent-orchestration/foundry-workflows-visual-orchestration.md",
"line": 33,
"claim": "Workflows er ett av tre agenttyper i Foundry: Prompt-based | Workflow | Hosted (preview, containeriserte agenter)",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/agent-orchestration/foundry-workflows-visual-orchestration.md#6",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/azure/foundry/agents/concepts/tool-catalog",
"evidence_quote": "The following table lists all built-in tools available in Foundry Agent Service. | Web search | Code Interpreter | Custom Code Interpreter (preview) | File Search | Azure AI Search | Azure Functions | Function calling | Image Generation (preview) | Browser Automation (preview) | Computer Use (preview) | Microsoft Fabric (preview) | SharePoint (preview)",
"reason": "The canonical tool catalog enumerates the built-in and custom tools (Code Interpreter, Azure AI Search, MCP, and Bing/Web search are present) but Key Vault is absent — in Foundry docs Key Vault is a secret-store connection, not a tool in the agent tool catalog, so one asserted list member does not exist.",
"file": "skills/ms-ai-engineering/references/agent-orchestration/foundry-workflows-visual-orchestration.md",
"line": 274,
"claim": "Verktøy tilgjengelige i agent-noder i workflows (samme katalog som enkelt-agenter): Code Interpreter, Bing Search, Azure AI Search, Key Vault, MCP-servere",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/agent-orchestration/foundry-workflows-visual-orchestration.md#8",
"judge_verdict": "not_grounded",
"rule": "R3",
"evidence_url": "https://learn.microsoft.com/azure/logic-apps/add-agent-action-create-run-workflow",
"evidence_quote": "Azure Logic Apps supports 1,400+ connectors and native, built-in data operations, so agents can integrate with many Microsoft and non-Microsoft services or products.",
"reason": "The docs attach the 1,400+ figure to connectors, not triggers, and state no '400+ enterprise connectors' count anywhere — the enterprise-connector category is described as a small set (SAP, IBM MQ, IBM 3270) with Salesforce classified as a Standard connector, so the claim's organizing units and figures do not match the source.",
"file": "skills/ms-ai-engineering/references/agent-orchestration/foundry-workflows-visual-orchestration.md",
"line": 306,
"claim": "Azure Logic Apps: 1400+ triggere (HTTP, Events, Schedule, Queues, SaaS) | 400+ enterprise-koblinger (SAP, Salesforce, AS2, EDI)",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/agent-orchestration/foundry-workflows-visual-orchestration.md#14",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/azure/foundry/observability/how-to/how-to-monitor-agents-dashboard?view=foundry",
"evidence_quote": "Use these definitions to interpret the dashboard: Token usage ... Latency ... Run success rate ... Evaluation metrics ... Red teaming results",
"reason": "The dashboard's canonical metric enumeration lists Token usage, Latency, Run success rate, Evaluation metrics, and Red teaming results — 'Error rate' is not among the dashboard metrics (it appears only in a note about onboarding external custom agents), so one asserted list member is absent.",
"file": "skills/ms-ai-engineering/references/agent-orchestration/foundry-workflows-visual-orchestration.md",
"line": 391,
"claim": "Agent Monitoring Dashboard-metrikker: Token usage | Latency | Run success rate | Error rate | Evaluation metrics",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-engineering/references/agent-orchestration/semantic-kernel-agents-implementation.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 8,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-engineering/agent-orchestration/semantic-kernel-agents-implementation.md#1",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/agent-framework/overview/agent-framework-overview",
"evidence_quote": "dotnet add package Microsoft.Agents.AI.Foundry --prerelease",
"reason": "The canonical overview and support pages state no '1.0' version, no GA status, and no 2026-04-03 date anywhere on Learn, and the live getting-started still instructs installing prerelease packages (--prerelease for .NET, 'pip install --pre agent-framework-foundry' in the migration guide; Go explicitly 'in public preview'), so the asserted GA/production-ready status is absent from the pages that would state it and contradicted by current prerelease packaging.",
"file": "skills/ms-ai-engineering/references/agent-orchestration/semantic-kernel-agents-implementation.md",
"line": 4,
"claim": "Microsoft Agent Framework 1.0 nådde GA 3. april 2026; produksjonsklart open-source rammeverk (.NET + Python)",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/agent-orchestration/semantic-kernel-agents-implementation.md#3",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/agent-framework/workflows/",
"evidence_quote": "Multi-Agent Orchestration: Built-in patterns for coordinating multiple AI agents, including sequential, concurrent, hand-off, and magentic.",
"reason": "The MAF workflows page lists only four built-in patterns (Group Chat is absent) and states no 'stable (GA) in MAF 1.0' status anywhere, while the cited Semantic Kernel orchestration page carries the banner 'Agent Orchestration features in the Agent Framework are in the experimental stage' — the five-pattern and GA-stability load-bearing parts both fail even though checkpointing/HITL are documented.",
"file": "skills/ms-ai-engineering/references/agent-orchestration/semantic-kernel-agents-implementation.md",
"line": 9,
"claim": "De fem orkestreringsmønstrene Sequential, Concurrent, Handoff, Group Chat og Magentic er stabile (GA) i MAF 1.0 med streaming, checkpointing, human-in-the-loop og pause/resume",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/agent-orchestration/semantic-kernel-agents-implementation.md#8",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/agent-framework/support/",
"evidence_quote": "",
"reason": "Learn pages confirm both are open-source with GitHub repositories but no learn.microsoft.com page states the MIT license for either Semantic Kernel or Agent Framework; the license terms live in the GitHub repos, which are outside the allowed source.",
"file": "skills/ms-ai-engineering/references/agent-orchestration/semantic-kernel-agents-implementation.md",
"line": 417,
"claim": "Semantic Kernel: MIT License | Microsoft Agent Framework: MIT License",
"disposition": "unsourced"
}
]
},
{
"file": "skills/ms-ai-engineering/references/api-management/apim-ai-gateway-overview.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 7,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-engineering/api-management/apim-ai-gateway-overview.md#2",
"judge_verdict": "not_grounded",
"rule": "R3",
"evidence_url": "https://learn.microsoft.com/azure/api-management/api-management-policies",
"evidence_quote": "Limit large language model API token usage | Prevents large language model (LLM) API usage spikes by limiting LLM tokens per calculated key.",
"reason": "The canonical policy reference's AI gateway and rate-limiting tables enumerate only llm-token-limit, llm-emit-token-metric, llm-semantic-cache-lookup/store and llm-content-safety — azure-openai-token-limit and azure-openai-emit-token-metric are absent, and their doc URLs now redirect to the llm-* pages (which cover OpenAI, Anthropic and Google Vertex APIs), so the claimed dual azure-openai-*/llm-* taxonomy describes a superseded policy split.",
"file": "skills/ms-ai-engineering/references/api-management/apim-ai-gateway-overview.md",
"line": 82,
"claim": "Token-policies i APIM: azure-openai-token-limit (Azure OpenAI) | llm-token-limit (alle LLM-er) | azure-openai-emit-token-metric | llm-emit-token-metric",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-engineering/references/api-management/apim-authentication-oauth-managed-identity.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 6,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-engineering/api-management/apim-authentication-oauth-managed-identity.md#2",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/ai-foundry/openai/how-to/role-based-access-control",
"evidence_quote": "Cognitive Services OpenAI User - Cognitive Services OpenAI Contributor - Cognitive Services Contributor - Cognitive Services Usages Reader",
"reason": "The canonical Azure OpenAI RBAC page enumerates the fourth role as 'Cognitive Services Usages Reader' (subscription-level quota viewing), not a generic 'Reader (lese)' role, and shows Cognitive Services Contributor is not 'full tilgang' (it cannot access quota or make inference API calls with Microsoft Entra ID) — two stated load-bearing parts fail, so the whole claim is not grounded.",
"file": "skills/ms-ai-engineering/references/api-management/apim-authentication-oauth-managed-identity.md",
"line": 94,
"claim": "RBAC-roller for Azure OpenAI: Cognitive Services OpenAI User (bruke deployments) | Cognitive Services OpenAI Contributor (opprette/administrere deployments) | Cognitive Services Contributor (full tilgang) | Reader (lese)",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-engineering/references/api-management/backend-pool-management.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 10,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-engineering/api-management/backend-pool-management.md#10",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/azure-monitor/reference/tables/apimanagementgatewayllmlog",
"evidence_quote": "ApiVersion, CompletionTokens, CorrelationId, DeploymentName, IsStreamCompletion, ModelName, OperationName, PromptTokens, Region, RequestId, RequestMessages, ResponseMessages, SequenceNumber, TimeGenerated, TotalTokens, Type",
"reason": "The canonical ApiManagementGatewayLlmLog schema page enumerates all columns and confirms TotalTokens, PromptTokens, and CompletionTokens, but no BackendUrl column exists in that enumeration — a stated, load-bearing column name a reader would copy into a KQL query is wrong, so the whole claim fails under R8/R2 despite the other parts holding.",
"file": "skills/ms-ai-engineering/references/api-management/backend-pool-management.md",
"line": 405,
"claim": "Log Analytics-tabellen ApiManagementGatewayLlmLog finnes med kolonner TotalTokens, PromptTokens, CompletionTokens, BackendUrl (og ApiManagementGatewayLogs med BackendResponseCode)",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-api-best-practices.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 12,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-engineering/azure-ai-services/ai-services-api-best-practices.md#1",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/azure/foundry/openai/supported-languages#error-handling",
"evidence_quote": "The following errors are automatically retired twice by default with a brief exponential backoff: Connection Errors, 408 Request Timeout, 429 Rate Limit, >=500 Internal Errors",
"reason": "The claim asserts Python, .NET and Go all retry up to 3 times, but the live page states the Python (and JS) OpenAI SDK retries twice by default — only the .NET client classes retry 'up to three more times' — so a load-bearing part is contradicted.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-api-best-practices.md",
"line": 87,
"claim": "Azure OpenAI SDK-er (Python, .NET, Go) retrier automatisk 408|429|500|502|503|504 — opptil 3 ganger med eksponentiell backoff",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/ai-services-api-best-practices.md#10",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/foundry/openai/concepts/provisioned-throughput-billing",
"evidence_quote": "With Reservations, you commit to payment for a fixed number of PTUs over a one-month or one-year term, and in return, you receive a discounted effective $/PTU/hr rate.",
"reason": "The claim asserts 1-year (~37%) and 3-year (~57%) commitment discounts, but the live billing page states reservation terms are one-month or one-year only — no 3-year term exists, and no discount percentages are documented ('The discount varies by model family and term length').",
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-api-best-practices.md",
"line": 627,
"claim": "Provisioned reservation-rabatter: 1-årig commitment ~37%|3-årig commitment ~57%",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/ai-services-api-best-practices.md#11",
"judge_verdict": "not_grounded",
"rule": "R3",
"evidence_url": "https://learn.microsoft.com/azure/foundry-classic/foundry-models/supported-languages",
"evidence_quote": "Azure AI Inference beta SDK is deprecated and will be retired on August 26, 2026. Switch to the generally available OpenAI/v1 API with a stable OpenAI SDK.",
"reason": "The claim's organizing frame — azure-ai-inference/Azure.AI.Inference/@azure/ai-inference as the Foundry-supporting SDKs — is superseded: the Azure AI Inference SDK family is deprecated (the migration table maps openai as the replacement for azure-ai-inference, retiring May 30, 2026) and Microsoft directs to the OpenAI SDKs, even though the Go openai-go part is still correct.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-api-best-practices.md",
"line": 482,
"claim": "SDK-er som støtter Microsoft Foundry: Python azure-ai-inference + openai (Azure-variant)|.NET Azure.AI.Inference + Azure.AI.OpenAI|JS/TS @azure/openai + @azure/ai-inference|Go github.com/openai/openai-go (med Azure endpoint)",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-cost-optimization.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 10,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-engineering/azure-ai-services/ai-services-cost-optimization.md#10",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/foundry/openai/how-to/provisioned-throughput-sizing",
"evidence_quote": "The following Fireworks on Microsoft Foundry models support both Global and US Data Zone provisioned throughput.",
"reason": "The claim asserts PTU is only for Azure OpenAI, but the live sizing page publishes provisioned-throughput deployment parameters for non-OpenAI Foundry Models (Llama-3.3-70B-Instruct, DeepSeek-R1, DeepSeek-V3-0324, and the Fireworks model family), so the load-bearing 'kun for Azure OpenAI' exclusivity is superseded even though PTU still doesn't apply to classic AI Services like Speech or Vision.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-cost-optimization.md",
"line": 177,
"claim": "PTU er kun for Azure OpenAI — ikke for øvrige Azure AI Services",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-enterprise-architecture.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 16,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-engineering/azure-ai-services/ai-services-enterprise-architecture.md#8",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/reliability/reliability-ai-search",
"evidence_quote": "Tier: Your service must be on the Basic tier or higher ... Number of replicas: Your service must have at least two replicas",
"reason": "The claim asserts Standard tier or higher plus minimum 3 replicas, but the live page states zone redundancy requires only Basic tier or higher and at least two replicas (3 replicas is the read-write SLA condition, not the zone-redundancy requirement), so two load-bearing values are contradicted even though the no-built-in-DR part holds.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-enterprise-architecture.md",
"line": 81,
"claim": "Azure AI Search zone redundancy krever Standard tier eller høyere + minimum 3 replicas; ingen built-in disaster recovery (krever manuell gjenoppbygging eller support-kontakt)",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/ai-services-enterprise-architecture.md#11",
"judge_verdict": "not_grounded",
"rule": "",
"evidence_url": "https://learn.microsoft.com/azure/ai-foundry/openai/concepts/models#model-summary-table-and-region-availability",
"evidence_quote": "| norwayeast | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ |",
"reason": "The claim says no Azure OpenAI region exists in Norway, but the canonical model/region availability table lists norwayeast with standard regional deployments of gpt-4.1, o3-mini, o1, gpt-4o and more (gpt-4o was available there well before 2026-02), directly refuting the non-existence claim.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-enterprise-architecture.md",
"line": 364,
"claim": "Ingen Azure OpenAI-region i Norge per 2026-02; nærmeste regioner er Sweden Central og West Europe",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/ai-services-enterprise-architecture.md#12",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/api-management/api-management-features",
"evidence_quote": "Multi-region deployment | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ✔️ | ❌ ... Scale units | ❌ (automatic scaling) | 1 | 2 | 10 | 4 | 10 | 12 per region | 30",
"reason": "The live tier comparison contradicts multiple load-bearing parts: multi-region deployment is Premium-only (the claim gives it to Standard), Basic allows 2 scale units (claim says 1 unit max), and Premium classic is 12 units per region; the per-tier call volumes are not stated on the page either.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-enterprise-architecture.md",
"line": 435,
"claim": "APIM-tiers: Developer (1M calls, ingen SLA) | Basic (1M calls, SLA, 1 unit max) | Standard (10M calls, multi-region, 4 units) | Premium (unlimited, multi-region, VNet, 10+ units)",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/ai-services-enterprise-architecture.md#13",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/api-management/genai-gateway-capabilities",
"evidence_quote": "APPLIES TO: All API Management tiers ... Currently, the backend circuit breaker isn't supported in the Consumption tier of API Management.",
"reason": "The circuit-breaker-not-in-Consumption part is confirmed on the backends page, but the second load-bearing part 'Standard tier minimum for AI gateway' is contradicted: the AI gateway page applies to all API Management tiers (with per-capability variation), so no Standard-minimum exists.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-enterprise-architecture.md",
"line": 440,
"claim": "APIM circuit breaker er ikke tilgjengelig i Consumption tier; Standard tier minimum for AI gateway",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/ai-services-enterprise-architecture.md#14",
"judge_verdict": "not_grounded",
"rule": "",
"evidence_url": "https://learn.microsoft.com/azure/foundry/openai/how-to/prompt-caching",
"evidence_quote": "Prompt caching is enabled by default for all supported models. There's no opt-out support for prompt caching.",
"reason": "The claim describes prompt caching for Azure OpenAI as a planned, not-yet-available feature, but the live page documents it as shipped, enabled by default, with discounted cache-read pricing and extended 24h retention on newer models — the status is superseded.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-enterprise-architecture.md",
"line": 427,
"claim": "Caching (prompt caching) for Azure OpenAI omtales som planlagt feature (ikke tilgjengelig ennå)",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-governance-compliance.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 12,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-engineering/azure-ai-services/ai-services-governance-compliance.md#6",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/azure/search/search-security-built-in",
"evidence_quote": "",
"reason": "The current data-residency section confirms object names (indexes, fields, indexers) appear in telemetry logs processed outside the selected region for Microsoft support, but no live learn.microsoft.com page states the claimed 1.5-year global retention period, so the specific retention value can be neither confirmed nor refuted.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-governance-compliance.md",
"line": 114,
"claim": "Telemetry logs (objektnavn som indexer, skillsets) lagres globalt i 1,5 år for Microsoft-support (unntak fra Geography-residency)",
"disposition": "unsourced"
},
{
"id": "ms-ai-engineering/azure-ai-services/ai-services-governance-compliance.md#7",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/azure/azure-monitor/logs/data-retention-configure",
"evidence_quote": "The default retention period of Analytics tables in a Log Analytics workspace is 30 days. You can change the default analytics period of Analytics tables up to two years by modifying the workspace-level data retention setting.",
"reason": "The claim asserts a 90-day Log Analytics default, but the live page states the default retention is 30 days (adjustable up to 730), contradicting the load-bearing default-value part even though the 30-730 range roughly holds.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-governance-compliance.md",
"line": 139,
"claim": "Diagnostic logs standard retention: 90 dager (Log Analytics default), justerbar 30-730 dager",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/ai-services-governance-compliance.md#9",
"judge_verdict": "not_grounded",
"rule": "",
"evidence_url": "https://learn.microsoft.com/azure/key-vault/managed-hsm/firmware-update",
"evidence_quote": "the HSM firmware for both Managed HSM and Azure Key Vault Premium are officially upgraded to a modern version validated to FIPS 140-3 level 3 standard",
"reason": "The claim asserts FIPS 140-2 Level 3 for HSM-backed Key Vault, but live docs state both Key Vault Premium and Managed HSM are now validated to FIPS 140-3 Level 3 (portfolio table lists both as FIPS 140-3 Level 3), so the certification version is superseded.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-governance-compliance.md",
"line": 220,
"claim": "Keys kan lagres i HSM-backed Key Vault for FIPS 140-2 Level 3",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/ai-services-governance-compliance.md#12",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/purview/dlp-microsoft-teams",
"evidence_quote": "Office 365 and Microsoft 365 E3 include DLP protection for SharePoint, OneDrive, and Exchange.",
"reason": "The claim's load-bearing part 'DLP Policies krever E5/F5-lisens' is contradicted — E3 includes DLP for SharePoint/OneDrive/Exchange (E5/F5 is only needed for Teams chat and Endpoint DLP), and the M365 license comparison also shows DSPM for AI included at E3/E5, undercutting the 'ikke i standard lisens' part, so the bundled licensing claim fails even though the audit-included and pay-as-you-go parts hold.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-governance-compliance.md",
"line": 557,
"claim": "Purview-lisensiering: DLP Policies krever E5/F5-lisens (eller pay-as-you-go) | Sensitivity Labels krever E3/E5 | DSPM for AI ikke i standard lisens (pay-as-you-go tilgjengelig) | Audit (Unified Log) inkludert",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-monitoring-logging.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 10,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-engineering/azure-ai-services/ai-services-monitoring-logging.md#3",
"judge_verdict": "not_grounded",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/azure/azure-monitor/logs/data-retention-configure",
"evidence_quote": "By default, all tables in a Log Analytics workspace retain data for 30 days, except for log tables with 90-day default retention.",
"reason": "The claim says Log Analytics retention is 90 days free then paid, but the page states the default is 30 days (31 days included in the ingestion price), with 90-day free retention applying only to specific tables (Usage, AzureActivity, Application Insights App* tables).",
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-monitoring-logging.md",
"line": 100,
"claim": "Log Analytics data retention: 90 dager gratis, deretter betalt",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/ai-services-monitoring-logging.md#7",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/foundry-classic/openai/monitor-openai-reference",
"evidence_quote": "Azure OpenAI Requests: Number of calls made to the Azure OpenAI API over a period of time ... `AzureOpenAIRequests`",
"reason": "Four of the six names (TokenTransaction, GeneratedTokens, ProcessedPromptTokens, ActiveTokens) match the canonical metric reference, but the request metric is named `AzureOpenAIRequests` (not `Requests`) and no `Http429` metric exists anywhere in the enumeration (429s surface via the StatusCode dimension), so two load-bearing copy-into-code names are wrong.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-monitoring-logging.md",
"line": 294,
"claim": "Azure OpenAI metric-navn: TokenTransaction | GeneratedTokens | ProcessedPromptTokens | ActiveTokens (PTU) | Requests | Http429",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/ai-services-monitoring-logging.md#8",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/azure-monitor/app/application-insights-faq",
"evidence_quote": "The default Pay-as-you-go Log Analytics pricing tier includes 5 GB per month of free data allowance per billing account.",
"reason": "The 5 GB/month free allowance is confirmed, but it is granted per billing account, not per subscription as the claim asserts - the scoping unit is a load-bearing part and it is contradicted.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-monitoring-logging.md",
"line": 394,
"claim": "Application Insights: 5 GB/måned gratis data ingestion per subscription",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/ai-services-monitoring-logging.md#10",
"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": "Telemetry offers insights into your agent by tracking: Logged messages and events sent to and from your agent, Topics to trigger during user conversations, Custom telemetry events that you can send from your topics",
"reason": "The canonical Copilot Studio telemetry page confirms the App Insights connection and conversation analytics but enumerates only messages/events, topics, and custom events - no LUIS intent recognition or QnA Maker query latency (those belong to Bot Framework SDK bots, and QnA Maker was retired 31 March 2025), so two load-bearing parts fail.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-monitoring-logging.md",
"line": 332,
"claim": "Copilot Studio-bottar kan kobles til Application Insights for conversation analytics, LUIS intent recognition performance og QnA Maker query latency",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-networking-security.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 15,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-engineering/azure-ai-services/ai-services-networking-security.md#8",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/azure/ai-services/cognitive-services-virtual-networks",
"evidence_quote": "",
"reason": "Neither the cited networking page nor any Microsoft Learn page found via search states a pricing-tier prerequisite (S0/Basic or higher, Free tier excluded) for private endpoints/IP firewall/service endpoints on AI Services, so the tier requirement can be neither confirmed nor refuted.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-networking-security.md",
"line": 404,
"claim": "Nettverkssikkerhet (private endpoints, IP firewall, service endpoints) krever Basic tier (S0) eller høyere; Free tier ikke støttet",
"disposition": "unsourced"
},
{
"id": "ms-ai-engineering/azure-ai-services/ai-services-networking-security.md#15",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/power-platform/admin/virtual-network-support-whitepaper",
"evidence_quote": "The new architecture eliminated costs and limitations associated with on-premises data gateways. ... a connector workload can access the target resource or endpoint inside the same virtual network.",
"reason": "The claim asserts Power Automate/Power Apps 'require' an on-premises data gateway and that custom connectors 'must' use the public endpoint, but live docs document Power Platform Virtual Network support (subnet delegation) letting connectors reach private-endpoint resources directly without a gateway — the 'krever'/'må' load-bearing parts describe a superseded state.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-networking-security.md",
"line": 316,
"claim": "Power Automate/Power Apps mot private endpoints krever On-premises data gateway; Azure Relay hybrid connection ikke støttet direkte; custom connector må bruke public endpoint med IP firewall",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-vs-foundry-tools-selection.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 16,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-engineering/azure-ai-services/ai-services-vs-foundry-tools-selection.md#2",
"judge_verdict": "not_grounded",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/azure/ai-services/personalizer/what-is-personalizer",
"evidence_quote": "Starting 20 September 2023 you won't be able to create new Personalizer resources. The Personalizer service is being retired 1 October 2026.",
"reason": "The claim states Personalizer IS retired, but the live page states it is BEING retired on 1 October 2026 (still ~2.5 months away), i.e. deprecated/retiring, a different status value.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-vs-foundry-tools-selection.md",
"line": 58,
"claim": "Personalizer er utgått (retired)",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/ai-services-vs-foundry-tools-selection.md#6",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/foundry/foundry-models/concepts/models-sold-directly-by-azure",
"evidence_quote": "| GPT-5.6 series | **NEW** `gpt-5.6-sol`, `gpt-5.6-terra`, `gpt-5.6-luna` |",
"reason": "The claim's load-bearing part 'GPT-5.5 nyest' is superseded: the live model page lists the GPT-5.6 series (2026-07-09) as the newest flagship series (and also a GPT-5.3 series the claim omits).",
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-vs-foundry-tools-selection.md",
"line": 145,
"claim": "Azure OpenAI modellserie 2026-06: GPT-5-generasjonen (GPT-5/5.1/5.2/5.4/5.5 + -mini/-nano/-codex) er flaggskip for reasoning/chat; GPT-5.5 nyest",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/ai-services-vs-foundry-tools-selection.md#12",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/azure/bot-service/file-format/bot-builder-qna-file-format",
"evidence_quote": "Azure AI QnA Maker will be retired on 31 March 2025. ... Custom question answering, a feature of Azure AI Language, is the updated version of the QnA Maker service.",
"reason": "The retirement part is correct (31 March 2025 has passed), but the load-bearing replacement part is wrong: the documented successor is custom question answering in Azure AI Language, not 'Language Understanding' (LUIS), which is itself a retired product.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-vs-foundry-tools-selection.md",
"line": 387,
"claim": "QnA Maker er utgått (retired); erstattes av Language Understanding for FAQ-bots",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/ai-services-vs-foundry-tools-selection.md#16",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/azure/foundry/agents/how-to/agent-365",
"evidence_quote": "Microsoft 365 licensing that supports Microsoft Agent 365. Agent 365 works best with Microsoft E5, and at least one user in your organization must have a qualifying Microsoft Agent 365 license such as Microsoft 365 Copilot.",
"reason": "The Azure-subscription part holds, but the load-bearing licensing threshold is stated differently: the docs require a qualifying Agent 365 license such as Microsoft 365 Copilot (works best with E5), not 'M365 E3 or higher' — an E3 license alone does not meet the documented requirement.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/ai-services-vs-foundry-tools-selection.md",
"line": 559,
"claim": "Agent-publisering til Agent 365 krever M365 E3 eller høyere (Azure subscription + M365 E3/E5)",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-engineering/references/azure-ai-services/azure-ai-vision-image-analysis.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 14,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-engineering/azure-ai-services/azure-ai-vision-image-analysis.md#9",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/ai-services/computer-vision/faq",
"evidence_quote": "The free (S0) tier only allows 20 transactions per minute. Upgrade to the S1 tier to get up to 20 transactions per second.",
"reason": "The 'up to 20 requests/second on the Standard (S1) tier' part is grounded, but the stated '10-20 requests/sekund per tier' range is contradicted for the free tier, which allows only 20 transactions per minute (about 0.33/s), and no Learn page states a 10 rps tier for Azure AI Vision.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/azure-ai-vision-image-analysis.md",
"line": 122,
"claim": "Rate limits: 10-20 requests/sekund per tier (opp til 20 requests/sekund per Standard tier)",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/azure-ai-vision-image-analysis.md#10",
"judge_verdict": "not_grounded",
"rule": "",
"evidence_url": "https://learn.microsoft.com/azure/ai-services/custom-vision-service/limits-and-quotas",
"evidence_quote": "Min labeled images per Tag, Classification (50+ recommended) | 5 | 5",
"reason": "The claim overstates a recommendation as a requirement: the Custom Vision limits page sets the enforced minimum at 5 labeled images per tag for classification and 15 for object detection (50+ recommended), while the 30-images figure appears in quickstarts only as guidance ('you should use at least 30 images per tag'), so Custom Vision does not require 30.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/azure-ai-vision-image-analysis.md",
"line": 154,
"claim": "Custom Vision krever trening med minimum 30 bilder per tag",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/azure-ai-vision-image-analysis.md#13",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/en-us/azure/ai-services/computer-vision/overview-image-analysis",
"evidence_quote": "To use the Image Analysis APIs, you must create your Azure Vision in Foundry Tools resource in a supported region. The Image Analysis features are available in the following regions:",
"reason": "The canonical region-availability enumeration for the Image Analysis features this file covers lists 13 regions (East US, West US, West US 2, France Central, North Europe, West Europe, Sweden Central, Switzerland North, Australia East, Southeast Asia, East Asia, Korea Central, Japan East) and neither Norway East nor Norway West appears, so the claimed Norway availability is absent from the enumerating page.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/azure-ai-vision-image-analysis.md",
"line": 271,
"claim": "Azure AI Vision er tilgjengelig i Norway East og Norway West",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/azure-ai-vision-image-analysis.md#14",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/azure/ai-services/computer-vision/faq",
"evidence_quote": "",
"reason": "The FAQ confirms the Standard tier is named S1 and quickstarts confirm an F0 free tier exists, but no fetched or searched learn.microsoft.com page states the 0-5000 free transactions/month figure, which lives only on the JS-rendered Azure pricing page.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/azure-ai-vision-image-analysis.md",
"line": 281,
"claim": "Azure AI Vision Free tier (F0): 0-5000 transaksjoner/måned gratis; Standard tier heter S1",
"disposition": "unsourced"
}
]
},
{
"file": "skills/ms-ai-engineering/references/azure-ai-services/azure-ai-vision-ocr-processing.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 13,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-engineering/azure-ai-services/azure-ai-vision-ocr-processing.md#8",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/ai-builder/prebuilt-text-recognition",
"evidence_quote": "",
"reason": "Learn confirms the AI Builder text recognition prebuilt model exists with OCR and Power Automate integration, but no Learn page states it uses Azure Vision OCR under the hood ('uses state-of-the-art optical character recognition (OCR)' names no engine), so the load-bearing engine attribution can be neither confirmed nor refuted.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/azure-ai-vision-ocr-processing.md",
"line": 201,
"claim": "AI Builder tilbyr en Text Recognition prebuilt model som bruker Azure Vision OCR under panseret (Power Automate-integrasjon)",
"disposition": "unsourced"
},
{
"id": "ms-ai-engineering/azure-ai-services/azure-ai-vision-ocr-processing.md#10",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/foundry/responsible-ai/computer-vision/ocr-data-privacy-security",
"evidence_quote": "Deletes data: The input data and results are deleted within 24 hours and not used for any other purpose.",
"reason": "The cited privacy page states input data and results are deleted within 24 hours, contradicting the claimed 48-hour retention (48 hours appears in docs only as the Operation-Location URL expiry, not data retention), so a load-bearing part is wrong.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/azure-ai-vision-ocr-processing.md",
"line": 248,
"claim": "OCR data retention: input-bilder og ekstrahert tekst lagres midlertidig i 48 timer (operation-location URL), deretter slettet automatisk; ingen permanent lagring av kundedata",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/azure-ai-vision-ocr-processing.md#12",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/azure/ai-services/document-intelligence/service-limits?view=doc-intel-4.0.0",
"evidence_quote": "",
"reason": "No learn.microsoft.com page states the monthly free allowances (5,000 transactions or 500 pages); the DI service-limits page defers 'monthly allowances' for F0 to the JS-rendered Azure pricing page, which is the expected source_silent case for pricing values.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/azure-ai-vision-ocr-processing.md",
"line": 272,
"claim": "Gratisnivåer: Azure Vision v4.0 Read OCR 5000 transaksjoner/måned gratis (Standard S1) | Document Intelligence Read 500 sider/måned gratis (Standard S0)",
"disposition": "unsourced"
}
]
},
{
"file": "skills/ms-ai-engineering/references/azure-ai-services/content-understanding-multimodal-analysis.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 12,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-engineering/azure-ai-services/content-understanding-multimodal-analysis.md#5",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/ai-services/content-understanding/video/elements",
"evidence_quote": "This model may miss transitions that are visually gradual.",
"reason": "Uniform keyframe sampling, minimum one per shot, deterministic shot output, and 'first shot always starts at 0 ms' all check out, but the load-bearing claim that the service detects gradual transitions is contradicted — the page explicitly warns the model may miss visually gradual transitions.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/content-understanding-multimodal-analysis.md",
"line": 150,
"claim": "Keyframes samples uniformt fra hver camera shot, minimum én per shot, deterministisk utvalg; cameraShotTimesMs angir startpunkt per shot (første shot starter alltid ved 0 ms); detekterer abrupte og gradvise overganger",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/content-understanding-multimodal-analysis.md#7",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/ai-services/content-understanding/pricing-explainer",
"evidence_quote": "You pay per input unit processed: Documents: Per 1,000 pages; Audio and Video: Per minute.",
"reason": "The capability parts hold (Video Indexer celebrity/observed-people/live-stream support confirmed on insights and overview pages; CU GA lacks face identification), but the load-bearing pricing contrast is wrong: CU's GA pricing charges per-page/per-minute content extraction meters plus token-based generative charges, so 'CU prises token-basert' vs VI's minute-based pricing describes a superseded (preview-era) billing frame.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/content-understanding-multimodal-analysis.md",
"line": 242,
"claim": "Content Understanding vs Video Indexer: CU har ikke face recognition (kun face description via Limited Access) og ikke real-time analyse; Video Indexer har celebrity + custom faces, live video streaming, observed people tracking (bounding boxes); CU prises token-basert, Video Indexer page/minute-basert",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/content-understanding-multimodal-analysis.md#11",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/azure/ai-services/content-understanding/tutorial/logic-apps",
"evidence_quote": "You can use the free pricing tier (F0) to try the service, then upgrade to a paid tier for production.",
"reason": "The claim asserts no free tier exists, but a live Learn tutorial states a free pricing tier (F0) can be used to try Content Understanding, directly contradicting the non-existence claim.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/content-understanding-multimodal-analysis.md",
"line": 482,
"claim": "Free tier er ikke tilgjengelig for Content Understanding; krever betalt Azure subscription",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/content-understanding-multimodal-analysis.md#12",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/en-us/python/api/overview/azure/ai-contentunderstanding-readme",
"evidence_quote": "Content Understanding operations are asynchronous long-running operations. The workflow is: 1. Begin Analysis - Start the analysis operation (returns immediately with an operation location) 2. Poll for Results - Poll the operation location until the analysis completes",
"reason": "The claim asserts a webhook-based async-completion feature exists in preview, but the canonical async-operation documentation (SDK readme, REST quickstart 'Poll the URL every 1–2 seconds until status is Succeeded') enumerates only Operation-Location polling and no CU page mentions webhooks at all — absence from the enumerating pages is evidence the feature does not exist.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/content-understanding-multimodal-analysis.md",
"line": 547,
"claim": "Webhooks for async completion er preview-feature, ikke GA (per Feb 2026)",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-engineering/references/azure-ai-services/document-intelligence-custom-models.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 11,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-engineering/azure-ai-services/document-intelligence-custom-models.md#7",
"judge_verdict": "not_grounded",
"rule": "R4",
"evidence_url": "https://learn.microsoft.com/en-us/azure/ai-services/document-intelligence/train/composed-models?view=doc-intel-4.0.0",
"evidence_quote": "The **2024-11-30 (GA)** implementation of the `model compose` operation replaces the implicit classification from the earlier versions with an explicit classification step and adds conditional routing. ... **Assigned custom model maximum expanded to 500**.",
"reason": "The claim describes the legacy behavior the compose how-to explicitly marks as 'only applies to v3.1 and previous versions': in current v4.0 the limit is 500 (not 200) and the automatic best-match classification is replaced by an explicitly trained classifier.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/document-intelligence-custom-models.md",
"line": 89,
"claim": "Composed models: kombiner opptil 200 custom models til én modell-ID; Document Intelligence klassifiserer dokumentet automatisk og velger best match",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/document-intelligence-custom-models.md#9",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/en-us/ai-builder/create-form-processing-model",
"evidence_quote": "**General documents**: Previously known as *unstructured*, this option is ideal for any kind of documents, especially when there's no set structure, or when the format is complex. ... This model is powerful, but has long training time. ... **Overlapping fields**: v4.0 supports overlapping fields in custom models",
"reason": "The canonical AI Builder page enumerates a 'General documents' (unstructured, long training time) model type alongside 'Fixed template documents' and lists v4.0 features (overlapping fields, signature detection, table confidence) that Document Intelligence documents as neural-only capabilities, contradicting the claim that AI Builder supports only template models.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/document-intelligence-custom-models.md",
"line": 250,
"claim": "AI Builder custom models (Document Processing) støtter kun template models, ikke neural (per januar 2026)",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/document-intelligence-custom-models.md#11",
"judge_verdict": "not_grounded",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/azure/ai-services/document-intelligence/train/custom-lifecycle?view=doc-intel-4.0.0",
"evidence_quote": "The model is configured to expire two years after its creation for all requests utilizing a GA API to build it.",
"reason": "No Microsoft Learn page states a 90-day storage period for Document Intelligence custom models; the lifecycle page states GA-trained models expire two years after creation, a differing value for the claimed model lifetime.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/document-intelligence-custom-models.md",
"line": 368,
"claim": "Custom models lagres i 90 dager uten kostnad",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-engineering/references/azure-ai-services/document-intelligence-prebuilt-models.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 12,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-engineering/azure-ai-services/document-intelligence-prebuilt-models.md#2",
"judge_verdict": "not_grounded",
"rule": "R3",
"evidence_url": "https://learn.microsoft.com/azure/ai-services/document-intelligence/language-support/prebuilt?view=doc-intel-4.0.0",
"evidence_quote": "The following tables list the available language and locale support by model and feature ... Model ID: prebuilt-bankStatement | English (United States) en-US",
"reason": "The canonical language-support page has no uniform 27-language figure and its per-model frame contradicts one: support ranges from en-US only (bankStatement, check, contract, tax, mortgage) to ~40 languages (invoice) to 100+ (thermal receipts), so 'prebuilt models support 27 languages' matches no current enumeration.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/document-intelligence-prebuilt-models.md",
"line": 30,
"claim": "Prebuilt-modellene støtter 27 språk",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/document-intelligence-prebuilt-models.md#4",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/azure/ai-services/document-intelligence/model-overview?view=doc-intel-4.0.0",
"evidence_quote": "prebuilt-check.us | ✓ | ✓",
"reason": "Six of the seven model IDs match the page's model analysis features table, but the bank-check model ID is prebuilt-check.us, not prebuilt-check as claimed - a copy-into-code SKU string, so one load-bearing part is wrong.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/document-intelligence-prebuilt-models.md",
"line": 40,
"claim": "Financial Services prebuilt-modeller: prebuilt-invoice | prebuilt-receipt | prebuilt-bankStatement | prebuilt-creditCard | prebuilt-check | prebuilt-contract | prebuilt-payStub.us",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/document-intelligence-prebuilt-models.md#5",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/azure/ai-services/document-intelligence/model-overview?view=doc-intel-4.0.0",
"evidence_quote": "prebuilt-marriageCertificate.us | ✓ | ✓",
"reason": "The tax and ID model IDs match (prebuilt-tax.us.w2/.1098/.1099/.1040 and prebuilt-tax.us all appear), but the marriage-certificate model ID is prebuilt-marriageCertificate.us, not prebuilt-marriageCertificate as claimed - one load-bearing SKU string is wrong.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/document-intelligence-prebuilt-models.md",
"line": 52,
"claim": "Identity & Tax prebuilt-modeller: prebuilt-idDocument | prebuilt-healthInsuranceCard.us | prebuilt-marriageCertificate | prebuilt-tax.us.w2 | prebuilt-tax.us.1098 | prebuilt-tax.us.1099 | prebuilt-tax.us.1040 | prebuilt-tax.us (unified)",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/document-intelligence-prebuilt-models.md#6",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/azure/ai-services/document-intelligence/model-overview?view=doc-intel-4.0.0",
"evidence_quote": "| Closing Disclosure | Extract closing, transaction costs, and loan details. | prebuilt-mortgage.us.closingDisclosure |",
"reason": "prebuilt-mortgage.us.1003/.1004/.1005/.1008 match the page, but the fifth model ID is prebuilt-mortgage.us.closingDisclosure, not prebuilt-mortgage.us.disclosure as claimed.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/document-intelligence-prebuilt-models.md",
"line": 65,
"claim": "US Mortgage prebuilt-modeller: prebuilt-mortgage.us.1003 | prebuilt-mortgage.us.1004 | prebuilt-mortgage.us.1005 | prebuilt-mortgage.us.1008 | prebuilt-mortgage.us.disclosure",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/document-intelligence-prebuilt-models.md#7",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/azure/ai-services/document-intelligence/model-overview?view=doc-intel-4.0.0",
"evidence_quote": "All the capabilities for the general document model are available in the layout model. The general model is no longer supported.",
"reason": "prebuilt-read and prebuilt-layout descriptions are grounded, but the page states the general document model (prebuilt-document) is no longer supported in v4.0 (its 2024-11-30 column reads 'Supported in layout model'), so presenting it as a current basic model contradicts the source.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/document-intelligence-prebuilt-models.md",
"line": 77,
"claim": "Grunnleggende modeller: prebuilt-read (OCR: tekst, linjer, ord, språkdeteksjon) | prebuilt-layout (tabeller, selection marks, seksjoner, valgfrie key-value pairs) | prebuilt-document (key-value pairs, tabeller, selection marks)",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/document-intelligence-prebuilt-models.md#8",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/ai-builder/credit-management",
"evidence_quote": "AI Builder credits coming from the AI Builder add-on (1,000,000 credits per add-on) and AI Builder credits coming from licenses with seeded capacity (like Power Automate premium, which brings 5,000 credits) are gathered at the tenant level.",
"reason": "The premium-feature part holds, but the 1M-credits part is contradicted: 1,000,000 credits come only from the purchased AI Builder capacity add-on, while Power Apps/Automate licenses seed just 250-5,000 credits (and seeded credits are removed November 1, 2026); the DI v3.1 part was not confirmed either.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/document-intelligence-prebuilt-models.md",
"line": 234,
"claim": "AI Builder bruker Document Intelligence v3.1 (ikke alltid v4.0); Premium-lisens påkrevd for AI Builder; 1M AI Builder credits inkludert i visse Power Apps/Automate-lisenser",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/document-intelligence-prebuilt-models.md#10",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/azure/ai-services/document-intelligence/service-limits?view=doc-intel-4.0.0",
"evidence_quote": "Analyze transactions Per Second limit | 1 | 15 (default value) ... Max number of pages (Analysis) | 2 | 2000",
"reason": "F0 2 pages/document, S0 2,000 pages, and S0 15 TPS all match, but the F0 rate limit is stated as 1 analyze transaction per second, not '20 calls/min' as claimed (and the 500 pages/month allowance appears only on the JS-rendered pricing page), so one load-bearing part is contradicted.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/document-intelligence-prebuilt-models.md",
"line": 391,
"claim": "Free (F0): 500 sider/måned, 2 sider per dokument, 20 calls/min | Standard (S0): 2 000 sider per dokument, 15 TPS",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/document-intelligence-prebuilt-models.md#11",
"judge_verdict": "not_grounded",
"rule": "",
"evidence_url": "https://learn.microsoft.com/azure/ai-services/document-intelligence/faq?view=doc-intel-4.0.0",
"evidence_quote": "Your data is then deleted 24 hours from the time that you submit an analyze request. If you would like the data deleted sooner, you can call the delete analyze response.",
"reason": "The documented standard retention is 24 hours (deletable sooner via the Delete Analyze Result API), not 30 days as claimed - the immediate-deletion part holds but the retention value is contradicted.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/document-intelligence-prebuilt-models.md",
"line": 355,
"claim": "Standard 30-dagers oppbevaring av dokumenter i Document Intelligence (kan slettes umiddelbart etter prosessering)",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/document-intelligence-prebuilt-models.md#12",
"judge_verdict": "not_grounded",
"rule": "",
"evidence_url": "https://learn.microsoft.com/azure/ai-services/document-intelligence/train/custom-lifecycle?view=doc-intel-4.0.0",
"evidence_quote": "The model is configured to expire two years after its creation for all requests utilizing a GA API to build it.",
"reason": "The retraining-need part holds, but the page states a uniform two-year expiration (also two years for preview-trained models), not the claimed '12-24 months' - no 12-month expiry exists, so the stated interval contradicts the documented value.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/document-intelligence-prebuilt-models.md",
"line": 488,
"claim": "Custom models utløper (slutter å virke) etter 12–24 måneder; krever retraining-schedule",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-engineering/references/azure-ai-services/language-services-custom-text-classification.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 12,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-engineering/azure-ai-services/language-services-custom-text-classification.md#7",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/microsoft-365/copilot/extensibility/overview",
"evidence_quote": "",
"reason": "No learn.microsoft.com page states whether custom text classification can or cannot be integrated directly into M365 Copilot or names the Copilot Studio/Power Automate workaround; this architectural claim is neither confirmed nor contradicted by any canonical page found.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/language-services-custom-text-classification.md",
"line": 266,
"claim": "Custom Text Classification kan ikke integreres direkte i M365 Copilot (Copilot bruker forhåndstrente modeller); workaround via Copilot Studio-bot eller Power Automate-flow",
"disposition": "unsourced"
},
{
"id": "ms-ai-engineering/azure-ai-services/language-services-custom-text-classification.md#8",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/azure/foundry/responsible-ai/language-service/data-privacy",
"evidence_quote": "Data sent in synchronous or asynchronous calls may be temporarily stored by Language for up to 48 hours only and is purged thereafter. ... To prevent this temporary storage of input data, the LoggingOptOut query parameter can be set accordingly.",
"reason": "The claim asserts no logging of runtime text and a 15-minute response cache, but the live page states input data may be temporarily stored for up to 48 hours by default (opt-out via LoggingOptOut); the no-logging and 15-minute parts are contradicted even though the LoggingOptOut parameter exists.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/language-services-custom-text-classification.md",
"line": 300,
"claim": "Tekst sendt til runtime-API logges ikke, men respons caches i 15 min (kan deaktiveres med loggingOptOut: true)",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/language-services-custom-text-classification.md#10",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/ai-services/language-service/custom-text-classification/service-limits",
"evidence_quote": "F | Training time | One hour per month ... F | Prediction Calls | 5,000 text records per month ... Count of deployments per project (free tier) | 0 | 1",
"reason": "The 5,000 text records/month and max 1 deployment parts are confirmed, but F0 training is limited to one hour per month (Unlimited applies only to the S tier), contradicting the claimed unlimited training; one load-bearing part wrong fails the whole claim.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/language-services-custom-text-classification.md",
"line": 366,
"claim": "Free tier (F0): 5000 text records per måned for Prediction API | trening ubegrenset | maks 1 deployment per prosjekt",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/language-services-custom-text-classification.md#11",
"judge_verdict": "not_grounded",
"rule": "R1",
"evidence_url": "https://learn.microsoft.com/en-us/azure/ai-services/language-service/custom-text-classification/service-limits",
"evidence_quote": "125,000 characters. You can send up to 25 documents as long as they collectively don't exceed 125,000 characters",
"reason": "The claim states a ceiling of 10 documents per request, but the live limits page states up to 25 documents per request; the stated maximum is superseded (upper-bound rule, no ratio leniency).",
"file": "skills/ms-ai-engineering/references/azure-ai-services/language-services-custom-text-classification.md",
"line": 376,
"claim": "Batch API: opptil 10 dokumenter per request",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/azure-ai-services/language-services-custom-text-classification.md#12",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/en-us/azure/ai-services/language-service/custom-text-classification/language-support",
"evidence_quote": "Custom text classification supports .txt files in the following languages:",
"reason": "The canonical language-support enumeration lists 93 languages (Afrikaans through Zulu), fewer than the claimed 100+; the multilingual option only covers these same supported languages, so the asserted floor is not met.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/language-services-custom-text-classification.md",
"line": 401,
"claim": "Custom Classification støtter 100+ språk (multilingual-funksjonalitet)",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-engineering/references/azure-ai-services/language-services-question-answering.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 15,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-engineering/azure-ai-services/language-services-question-answering.md#12",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/azure/ai-services/language-service/question-answering/concepts/azure-resources",
"evidence_quote": "",
"reason": "The cited page only states that CQA's management and prediction services are colocated in the same region; neither it nor any Learn page found via search states (or contradicts) that the Azure AI Search resource must be in the same region as the Language resource or that cross-region replication is disallowed.",
"file": "skills/ms-ai-engineering/references/azure-ai-services/language-services-question-answering.md",
"line": 388,
"claim": "Cross-region replication er ikke tillatt i CQA; Azure AI Search må være i samme region som Language resource",
"disposition": "unsourced"
},
{
"id": "ms-ai-engineering/azure-ai-services/language-services-question-answering.md#14",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "",
"evidence_quote": "",
"reason": "The F0 5,000-text-records/month allowance and S-tier unlimited hosted calls are Azure pricing-page (JS-rendered) values; no learn.microsoft.com page found states them (the CQA limits page only notes F0 upload is limited to three files).",
"file": "skills/ms-ai-engineering/references/azure-ai-services/language-services-question-answering.md",
"line": 465,
"claim": "Language resource med CQA: Free (F0) 5000 text records/måned | Standard (S) ubegrensede hosted calls",
"disposition": "unsourced"
}
]
},
{
"file": "skills/ms-ai-engineering/references/mlops-genaiops/prompt-flow-production-deployment.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "pass",
"claim_count": 13,
"verified": "2026-07-18",
"verified_by": "judge-v3.1",
"flags": []
},
{
"file": "skills/ms-ai-engineering/references/rag-architecture/citation-tracking.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 10,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-engineering/rag-architecture/citation-tracking.md#3",
"judge_verdict": "not_grounded",
"rule": "",
"evidence_url": "https://learn.microsoft.com/azure/foundry-classic/openai/references/on-your-data",
"evidence_quote": "Azure OpenAI On Your Data is deprecated and approaching retirement. The service will be retired on October 14, 2026",
"reason": "The claim asserts 'On Your Data' with citation annotations is GA, but the live page states the service is deprecated and retires October 14, 2026 with Microsoft recommending migration to Foundry Agent Service — a superseded status a reader would act on.",
"file": "skills/ms-ai-engineering/references/rag-architecture/citation-tracking.md",
"line": 26,
"claim": "Classic RAG (Azure AI Search + Azure OpenAI, 'On Your Data' med citation annotations) er GA",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/rag-architecture/citation-tracking.md#6",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/ai-foundry/concepts/evaluation-evaluators/rag-evaluators",
"evidence_quote": "Groundedness Pro uses the Azure AI Content Safety service and returns a boolean result instead of a numeric score",
"reason": "The AACS-delivery part holds, but the load-bearing 'threshold-based scoring' part is contradicted: the live page reserves 'Pass/Fail based on threshold (1-5 scale)' for the other evaluators and states Groundedness Pro returns a boolean True/False instead of a numeric score.",
"file": "skills/ms-ai-engineering/references/rag-architecture/citation-tracking.md",
"line": 213,
"claim": "Groundedness Pro: threshold-basert scoring, leveres av Azure AI Content Safety (per evaluerings-kall)",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-engineering/references/rag-architecture/rag-context-windows.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 11,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-engineering/rag-architecture/rag-context-windows.md#10",
"judge_verdict": "not_grounded",
"rule": "R3",
"evidence_url": "https://learn.microsoft.com/en-us/azure/foundry/openai/quotas-limits",
"evidence_quote": "Quota values for global batch are represented in terms of enqueued tokens. ... Global batch: | Model | Enterprise and MCA-E | Default | Monthly credit card-based subscriptions | MSDN subscriptions | Azure for Students, free trials | `gpt-4.1` | 5B | 200M | 50M | 90K | N/A | ... `gpt-4o` | 5B | 200M | 50M | 90K | N/A",
"reason": "The live page organizes batch quota by subscription/agreement type (Enterprise/MCA-E, Default, credit card, MSDN) as enqueued tokens, not per Quota Tier per month, and global batch for gpt-4.1 and gpt-4o is 5B/200M/50M — the claim's ~500M/~30M figures match only the separate Data zone batch table, so both the organizing frame and the headline values are wrong.",
"file": "skills/ms-ai-engineering/references/rag-architecture/rag-context-windows.md",
"line": 339,
"claim": "Azure OpenAI Batch quota (per Quota Tier): gpt-4.1 ~500M tokens/month (upper tier), ~30M (lower tier); gpt-4o ~500M (upper), ~30M (lower)",
"disposition": "outdated"
},
{
"id": "ms-ai-engineering/rag-architecture/rag-context-windows.md#11",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/azure/foundry-classic/how-to/flow-develop",
"evidence_quote": "",
"reason": "No learn.microsoft.com page states that Prompt Flow shows an estimated token count per node in the flow: the develop-flow page documents a trace view with duration and token cost of the flow (per-node detail is duration), and per-node token_consumption metrics exist only in Application Insights for deployed flows, so the claimed in-flow per-node estimate can neither be confirmed nor refuted.",
"file": "skills/ms-ai-engineering/references/rag-architecture/rag-context-windows.md",
"line": 273,
"claim": "Prompt Flow (Microsoft Foundry) shows estimated token count per node in the flow",
"disposition": "unsourced"
}
]
},
{
"file": "skills/ms-ai-governance/references/responsible-ai/content-safety-implementation.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 12,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-governance/responsible-ai/content-safety-implementation.md#1",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/azure/ai-services/content-safety/whats-new",
"evidence_quote": "Protected material detection for code (preview)",
"reason": "The claim marks Protected Material Code as GA, but What's new and the current quickstart both title it '(preview)' with no GA announcement (the Aug 2024 GA covered only Prompt Shields and Protected Material for text), so one load-bearing status in the bundle is wrong while the other statuses match.",
"file": "skills/ms-ai-governance/references/responsible-ai/content-safety-implementation.md",
"line": 37,
"claim": "Content Safety feature statuses: Analyze Text GA | Analyze Image GA | Prompt Shields GA | Groundedness Detection Preview | Protected Material Text GA | Protected Material Code GA | Custom Categories (Standard) Preview | Custom Categories (Rapid) Preview | Blocklists GA",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/content-safety-implementation.md#2",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/ai-services/content-safety/overview",
"evidence_quote": "Maximum length for grounding sources: 55,000 characters (per API call). Maximum text and query length: 7,500 characters.",
"reason": "The claim states 'Groundedness query+sources max 55K chars', but the live page gives 55,000 characters to grounding sources only and caps text/query separately at 7,500 characters, so that load-bearing limit is misstated even though the other limits (Analyze Text 10K, image 4MB JPEG/PNG/GIF/BMP/TIFF/WEBP, Prompt Shields 10K, PM-Text min 110) match.",
"file": "skills/ms-ai-governance/references/responsible-ai/content-safety-implementation.md",
"line": 37,
"claim": "Content Safety input limits: Analyze Text max 10K chars | Analyze Image JPEG/PNG/GIF/BMP/TIFF/WEBP max 4MB | Prompt Shields text max 10K chars | Groundedness query+sources max 55K chars | Protected Material Text min 110 chars",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/content-safety-implementation.md#6",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/microsoft-copilot-studio/knowledge-copilot-studio",
"evidence_quote": "The moderation levels range from **Lowest** to **Highest**. The lowest level generates the most answers, but they might contain harmful content. The highest level of content moderation generates fewer answers, and applies a stricter filter to restrict harmful content. The default moderation level is **High**.",
"reason": "The claim says moderation 'uses Azure OpenAI deployment settings' and that custom blocklists can be enabled in Agent Settings, but the live docs show content moderation is configured inside Copilot Studio itself (agent-, topic-, and prompt-level Lowest-Highest levels) and the Agent settings enumeration has a Moderation level dropdown with no custom-blocklist option, so two load-bearing parts are wrong (only the no-per-category-severity part holds).",
"file": "skills/ms-ai-governance/references/responsible-ai/content-safety-implementation.md",
"line": 239,
"claim": "Copilot Studio cannot (per feb 2026) configure severity levels per category - uses Azure OpenAI deployment settings; custom blocklists can be enabled in Agent Settings",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/content-safety-implementation.md#7",
"judge_verdict": "not_grounded",
"rule": "R3",
"evidence_url": "https://learn.microsoft.com/ai-builder/prebuilt-azure-openai",
"evidence_quote": "This feature is deprecated and isn't visible anymore.",
"reason": "AI Builder's Text generation model is deprecated and replaced by prompt builder, where per the prompts FAQ 'Makers can configure the content moderation level for harmful content only' - so the claim's frame (no configuration options, implicit Medium+High default) describes a replaced world and its 'no configuration' part is contradicted for the current mechanism.",
"file": "skills/ms-ai-governance/references/responsible-ai/content-safety-implementation.md",
"line": 243,
"claim": "AI Builder Text generation uses Azure OpenAI with content filtering enabled by default and no configuration options (default Medium+High block)",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/content-safety-implementation.md#8",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/microsoft-365/copilot/harmful-content-protection-copilot-chat",
"evidence_quote": "For these special use cases, your organization's Microsoft 365 Apps administrator can enable a specific group of users to adjust harmful content protection settings in their Copilot Chat experiences.",
"reason": "The claim's load-bearing 'not customer-configurable' is contradicted by the live page: admins can policy-enable users to toggle harmful content protection off in Copilot Chat (and the privacy doc says Microsoft 'offers certain content filtering controls for admins and users'), even though severity-level sliders indeed do not exist.",
"file": "skills/ms-ai-governance/references/responsible-ai/content-safety-implementation.md",
"line": 259,
"claim": "Microsoft 365 Copilot has its own content filtering policies that are not customer-configurable (severity levels cannot be adjusted; Microsoft-managed)",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/content-safety-implementation.md#9",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/en-us/azure/ai-services/content-safety/overview",
"evidence_quote": "To use the Content Safety APIs, you must create your Azure AI Content Safety resource in a supported region. Currently, the Content Safety features are available in the following Azure regions with different API versions:",
"reason": "The canonical region-availability table on the cited page lists West Europe but contains no Norway East row for any Content Safety API, and no Learn page found states Content Safety availability in Norway East (even via Azure OpenAI), so the Norway East part fails the enumeration check; the in-region/no-US-transfer part alone is supported by the data-privacy page.",
"file": "skills/ms-ai-governance/references/responsible-ai/content-safety-implementation.md",
"line": 275,
"claim": "Azure AI Content Safety available in West Europe and Norway East (via Azure OpenAI); models run in EU (no data transfer to USA)",
"disposition": "outdated"
},
{
"id": "ms-ai-governance/responsible-ai/content-safety-implementation.md#10",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/azure/ai-services/language-service/personally-identifiable-information/concepts/entity-categories",
"evidence_quote": "To retrieve this entity type, specify **NOIdentityNumber** in the **piiCategories** request parameter. If detected, the entity appears in the **PII** response payload.",
"reason": "The PII filter page links this entity-categories list as 'the complete list of supported personal data entity types' and lists 'National ID numbers (50+ countries)' including a dedicated Norway Identity Number (NOIdentityNumber) type, contradicting the claim's load-bearing assertion that the Norwegian national ID format is not officially supported (the detection-in-completions and block/redacted_text parts do hold).",
"file": "skills/ms-ai-governance/references/responsible-ai/content-safety-implementation.md",
"line": 270,
"claim": "Azure AI Content Safety has PII detection for completions (names, addresses; Norwegian national ID format not officially supported), configurable to block or mask PII in LLM output",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-security/references/ai-security-engineering/ai-prompt-shield-network.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 12,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-security/ai-security-engineering/ai-prompt-shield-network.md#1",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/entra/fundamentals/whats-new-archive",
"evidence_quote": "General Availability - Protect enterprise GenAI applications with Prompt Injection Protection",
"reason": "The Content Safety Prompt Shields GA part holds, but the Entra whats-new archive (November 2025) announces General Availability of the AI Gateway prompt protection capability (now named Prompt Injection Protection), contradicting the claim's Preview status.",
"file": "skills/ms-ai-security/references/ai-security-engineering/ai-prompt-shield-network.md",
"line": 6,
"claim": "Two separate products: Azure AI Content Safety Prompt Shields is GA; AI Gateway Prompt Shield via Global Secure Access is Preview",
"disposition": "outdated"
},
{
"id": "ms-ai-security/ai-security-engineering/ai-prompt-shield-network.md#3",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/azure/ai-services/content-safety/overview",
"evidence_quote": "Prompt Shields API: Maximum prompt length: 10K characters. Up to five documents with a total of 10K characters.",
"reason": "Prompt limit 10K and max 5 documents hold, but the claim's 'each document max 10 000 chars' contradicts the source, which caps all documents at 10K characters combined, not per document.",
"file": "skills/ms-ai-security/references/ai-security-engineering/ai-prompt-shield-network.md",
"line": 139,
"claim": "Prompt Shields input limits: userPrompt max 10 000 chars | documents array max 5 documents per request | each document max 10 000 chars",
"disposition": "outdated"
},
{
"id": "ms-ai-security/ai-security-engineering/ai-prompt-shield-network.md#4",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/entra/global-secure-access/how-to-ai-prompt-injection-protection",
"evidence_quote": "Prompt Injection Protection is preconfigured with custom extractors for the following models: ChatGPT, Claude, Cohere, Deepseek, Gemini, Grok, Meta AI, Mistral, Perplexity, Pi, and Qwen.",
"reason": "The canonical enumeration does not include 'Microsoft Copilot' and lists 'Meta AI' rather than 'Meta Llama' (and now also includes Deepseek, Gemini, Perplexity), so the claim's stated extractor list disagrees with the current page.",
"file": "skills/ms-ai-security/references/ai-security-engineering/ai-prompt-shield-network.md",
"line": 180,
"claim": "AI Gateway Prompt Shield has pre-configured extractors for: Microsoft Copilot, OpenAI ChatGPT, Anthropic Claude, Meta Llama, xAI Grok, Mistral, Cohere, Inflection Pi, Alibaba Qwen, plus custom JSON-based LLMs (custom URL + JSON path)",
"disposition": "outdated"
},
{
"id": "ms-ai-security/ai-security-engineering/ai-prompt-shield-network.md#5",
"judge_verdict": "not_grounded",
"rule": "R1",
"evidence_url": "https://learn.microsoft.com/entra/global-secure-access/how-to-ai-prompt-injection-protection",
"evidence_quote": "Prompt Injection Protection supports prompts up to 64,000 characters. Anything longer is truncated.",
"reason": "Text-only and JSON-only limitations still hold, but the claimed 10 000-char ceiling is superseded — the live page states 64,000 characters, and Gemini (the claim's example of an unsupported app) is now a preconfigured extractor.",
"file": "skills/ms-ai-security/references/ai-security-engineering/ai-prompt-shield-network.md",
"line": 193,
"claim": "AI Gateway Prompt Shield limitations: text prompts only (no files); JSON-based GenAI apps only (not URL-encoded, e.g. Gemini); max 10 000 chars per prompt (longer prompts truncated)",
"disposition": "outdated"
},
{
"id": "ms-ai-security/ai-security-engineering/ai-prompt-shield-network.md#11",
"judge_verdict": "not_grounded",
"rule": "R7",
"evidence_url": "https://learn.microsoft.com/azure/ai-services/content-safety/quickstart-groundedness",
"evidence_quote": "The current version is: api-version=2024-09-15-preview. Example: <endpoint>/contentsafety/text:detectGroundedness?api-version=2024-09-15-preview",
"reason": "The endpoint path matches, but the load-bearing api-version string is wrong — the canonical quickstart states 2024-09-15-preview as the current version, not 2024-09-01.",
"file": "skills/ms-ai-security/references/ai-security-engineering/ai-prompt-shield-network.md",
"line": 309,
"claim": "Groundedness endpoint: POST {endpoint}/contentsafety/text:detectGroundedness?api-version=2024-09-01",
"disposition": "outdated"
},
{
"id": "ms-ai-security/ai-security-engineering/ai-prompt-shield-network.md#12",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/entra/fundamentals/whats-new-archive",
"evidence_quote": "General Availability - Protect enterprise GenAI applications with Prompt Injection Protection",
"reason": "The license-inclusion part is supported (the feature's prerequisites require a Microsoft Entra Internet Access license), but the stated '(Preview)' status is contradicted — the November 2025 Entra release notes announce General Availability.",
"file": "skills/ms-ai-security/references/ai-security-engineering/ai-prompt-shield-network.md",
"line": 428,
"claim": "AI Gateway Prompt Shield (Preview) is included in the Microsoft Entra Internet Access license (licensed per user/month)",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-security/references/ai-security-engineering/entra-agent-id-zero-trust.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 13,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-security/ai-security-engineering/entra-agent-id-zero-trust.md#1",
"judge_verdict": "not_grounded",
"rule": "R4",
"evidence_url": "https://learn.microsoft.com/entra/fundamentals/whats-new",
"evidence_quote": "The Microsoft Entra Agent ID platform is now generally available.",
"reason": "The claim asserts Public Preview status, but the April 2026 release notes announce general availability and the live product page now states 'Agent ID is available for all Microsoft Entra customers' with no preview disclaimer.",
"file": "skills/ms-ai-security/references/ai-security-engineering/entra-agent-id-zero-trust.md",
"line": 5,
"claim": "Microsoft Entra Agent ID har status Public Preview (utvidet etter Ignite 2025)",
"disposition": "outdated"
},
{
"id": "ms-ai-security/ai-security-engineering/entra-agent-id-zero-trust.md#3",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/microsoft-copilot-studio/admin-use-entra-agent-identities",
"evidence_quote": "Starting July 2026, all new agents must have Microsoft Entra Agent IDs, and you can no longer opt out of automatic agent identity creation.",
"reason": "The blueprint name and Blueprint ID 25664c89-cea5-4ab6-b924-a54fd8a19ae0 match, but the load-bearing 'preview, activated per environment in Power Platform Admin Center' framing is superseded - creation is now automatic and mandatory with the per-environment opt-out removed.",
"file": "skills/ms-ai-security/references/ai-security-engineering/entra-agent-id-zero-trust.md",
"line": 20,
"claim": "Copilot Studio (preview, aktiveres per miljø i Power Platform Admin Center) oppretter automatisk Entra Agent ID for hver ny agent, knyttet til Microsoft Copilot Studio agent identity blueprint med Blueprint ID 25664c89-cea5-4ab6-b924-a54fd8a19ae0",
"disposition": "outdated"
},
{
"id": "ms-ai-security/ai-security-engineering/entra-agent-id-zero-trust.md#10",
"judge_verdict": "not_grounded",
"rule": "R4",
"evidence_url": "https://learn.microsoft.com/en-us/microsoft-copilot-studio/admin-use-entra-agent-identities",
"evidence_quote": "Previously, you could opt out of Entra Agent ID at the environment level. Starting July 2026, all new agents must have Microsoft Entra Agent IDs, and you can no longer opt out of automatic agent identity creation.",
"reason": "The claim describes the opt-out as currently existing with mandatory enforcement coming 'in the future', but the current page states the opt-out is already removed and Agent IDs are mandatory for all new agents as of July 2026.",
"file": "skills/ms-ai-security/references/ai-security-engineering/entra-agent-id-zero-trust.md",
"line": 322,
"claim": "Opt-out fra Entra Agent Identity per miljø i Copilot Studio er midlertidig — Microsoft vil gjøre det obligatorisk for alle nye agenter i fremtiden",
"disposition": "outdated"
},
{
"id": "ms-ai-security/ai-security-engineering/entra-agent-id-zero-trust.md#11",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/entra/fundamentals/whats-new",
"evidence_quote": "The Microsoft Entra Agent ID platform is now generally available.",
"reason": "The status table's load-bearing 'Public Preview (Frontier program)' frame is superseded: the Agent ID platform reached GA in April 2026, the live product page lists Agent 365/E5/E7/P1-P2 licensing with no Frontier requirement, and the Entra Agent Registry blades were retired May 1, 2026 in the consolidation into Microsoft Agent 365 - even though the AI Prompt Shield, App Service/Functions, and Teams Developer Portal items individually check out.",
"file": "skills/ms-ai-security/references/ai-security-engineering/entra-agent-id-zero-trust.md",
"line": 439,
"claim": "Statustabell: Entra Agent ID kjerne/Agent Registry/Identity Protection for agenter/Global Secure Access for agenter = Public Preview (Frontier-program); Foundry-integrasjon Public Preview (alle Foundry-brukere); Conditional Access for agenter Public Preview; AI Prompt Shield nytt via Entra Internet Access; App Service/Azure Functions agent identity og Teams Developer Portal agent blueprints nye",
"disposition": "outdated"
},
{
"id": "ms-ai-security/ai-security-engineering/entra-agent-id-zero-trust.md#12",
"judge_verdict": "not_grounded",
"rule": "R3",
"evidence_url": "https://learn.microsoft.com/entra/agent-id/identity-professional/microsoft-entra-agent-identities-for-ai-agents",
"evidence_quote": "Extending Microsoft Entra security features to agents requires Microsoft 365 E7 (includes Agent 365 and Microsoft Entra Suite) or Microsoft 365 E5 paired with a Microsoft Agent 365 license.",
"reason": "The Frontier-program + M365 Copilot licensing frame has been replaced: the live page states full Agent ID security functionality requires Agent 365/E5/E7 (or P1/P2 standalone options) with no mention of Frontier enrollment or the Copilot Frontier admin-center toggle, which now applies only to pre-release Agent 365 features.",
"file": "skills/ms-ai-security/references/ai-security-engineering/entra-agent-id-zero-trust.md",
"line": 452,
"claim": "Full Entra Agent ID-funksjonalitet krever deltakelse i Microsoft Frontier-programmet og M365 Copilot-lisens (aktiveres via M365 admin center → Copilot → Settings → User access → Copilot Frontier); Foundry-integrert agentidentitet er tilgjengelig for alle Foundry-brukere uten Frontier",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-security/references/ai-security-engineering/zero-trust-ai-services.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 11,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-security/ai-security-engineering/zero-trust-ai-services.md#2",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/entra/identity/managed-identities-azure-resources/managed-identity-best-practice-recommendations",
"evidence_quote": "This means that it can take several hours for changes to a managed identity's group or role membership to take effect.",
"reason": "The 24-hour token cache and hours-long propagation parts are confirmed, but the page explicitly treats group AND app-role membership identically (both are claims in the cached token, same several-hours delay, and it recommends user-assigned MI with direct permissions for fast changes), contradicting the load-bearing part that App Roles propagate faster than groups.",
"file": "skills/ms-ai-security/references/ai-security-engineering/zero-trust-ai-services.md",
"line": 128,
"claim": "Managed Identity tokens are cached up to 24 hours; group/role membership changes can take hours to propagate; App Roles propagate faster than groups",
"disposition": "outdated"
},
{
"id": "ms-ai-security/ai-security-engineering/zero-trust-ai-services.md#3",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/azure/foundry/how-to/navigate-from-classic",
"evidence_quote": "Microsoft's AI Platform has evolved from Azure AI Studio -> Azure AI Foundry -> to Microsoft Foundry (current). Similarly, our AI services portfolio evolved with the platform from Azure Cognitive Services -> Azure AI Services -> to Foundry Tools (current).",
"reason": "The avoid-API-keys recommendation, kind=AIServices, and DefaultAzureCredential pattern are grounded, but the naming part 'formerly Foundry Tools' is contradicted: Foundry Tools is the CURRENT name of the AI services portfolio (formerly Azure AI Services), not a former name of the Microsoft Foundry resource (whose lineage is Azure AI Studio -> Azure AI Foundry -> Microsoft Foundry).",
"file": "skills/ms-ai-security/references/ai-security-engineering/zero-trust-ai-services.md",
"line": 130,
"claim": "Microsoft now explicitly recommends avoiding API keys for Azure AI Services in production; Microsoft Foundry resource (formerly 'Foundry Tools', kind=AIServices) uses the same DefaultAzureCredential pattern across all AI services",
"disposition": "outdated"
},
{
"id": "ms-ai-security/ai-security-engineering/zero-trust-ai-services.md#4",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/entra/identity/conditional-access/concept-continuous-access-evaluation",
"evidence_quote": "The goal for critical event evaluation is for response to be near real time, but latency of up to 15 minutes might be observed because of event propagation time; however, IP locations policy enforcement is instant.",
"reason": "The 28-hour CAE token lifetime and 1-hour standard token parts are confirmed on the page (and the MI 24-hour cache on the MI best-practices page), but the load-bearing part 'critical events revoke in seconds' is contradicted: the page states near real time with up to 15 minutes latency and 'within minutes after a critical event', not seconds.",
"file": "skills/ms-ai-security/references/ai-security-engineering/zero-trust-ai-services.md",
"line": 149,
"claim": "CAE token lifetime: up to 28 hours (vs standard access token 1 hour, Managed Identity refresh token 24 hours); critical events revoke in seconds",
"disposition": "outdated"
},
{
"id": "ms-ai-security/ai-security-engineering/zero-trust-ai-services.md#5",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/entra/identity/conditional-access/concept-continuous-access-evaluation-strict-enforcement",
"evidence_quote": "Standard (Default) | Suitable for all topologies | A short-lived token is issued only if Microsoft Entra ID detects an allowed IP address. Otherwise, access is blocked",
"reason": "The move to Conditional Access session controls is confirmed ('The CAE setting moved to Conditional Access.'), but the option set is wrong: the documented options are Disable (session control) and Strictly enforce location policies, with the default enforcement mode named 'Standard (Default)' - no option named 'Basic' exists on the session-control or strict-enforcement pages.",
"file": "skills/ms-ai-security/references/ai-security-engineering/zero-trust-ai-services.md",
"line": 152,
"claim": "From 2025 CAE is configured via Conditional Access policies (Session controls) instead of a separate toggle; options Disabled, Basic (default), Strict",
"disposition": "outdated"
},
{
"id": "ms-ai-security/ai-security-engineering/zero-trust-ai-services.md#6",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/en-us/entra/identity/conditional-access/concept-continuous-access-evaluation",
"evidence_quote": "Continuous access evaluation is also available in Azure Government tenants (GCC High and DOD) for Exchange Online.",
"reason": "Preview status, per-policy activation, and near-realtime blocking are confirmed on the strict-enforcement page, but 'now supports Azure Government clouds' is absent from both canonical pages - the only Government-availability statement covers CAE generally and only for Exchange Online, so the claimed strict-enforcement Government support is not documented.",
"file": "skills/ms-ai-security/references/ai-security-engineering/zero-trust-ai-services.md",
"line": 154,
"claim": "Strict Location Enforcement is Preview, activated per Conditional Access policy, blocks tokens used outside approved network locations in near-realtime, and now supports Azure Government clouds",
"disposition": "outdated"
},
{
"id": "ms-ai-security/ai-security-engineering/zero-trust-ai-services.md#9",
"judge_verdict": "not_grounded",
"rule": "",
"evidence_url": "https://learn.microsoft.com/power-platform/admin/vnet-support-overview",
"evidence_quote": "| Dataverse | Dataverse plug-ins | Generally available |",
"reason": "The cited vnet-data-gateway page returns 404; the live Virtual Network support overview states Dataverse (plug-ins) Virtual Network support is Generally available, contradicting the claimed 'Private Preview (Q1 2026)' status.",
"file": "skills/ms-ai-security/references/ai-security-engineering/zero-trust-ai-services.md",
"line": 606,
"claim": "Virtual Network Integration for Dataverse is in Private Preview (Q1 2026)",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-security/references/cost-optimization/azure-ai-foundry-cost-governance.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 9,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-security/cost-optimization/azure-ai-foundry-cost-governance.md#7",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/azure/foundry/how-to/navigate-from-classic",
"evidence_quote": "| Resource type | Azure OpenAI + Hub | Foundry Resource | Single `AIServices` kind with child projects. | | AI services | Azure AI Services | Foundry Tools | Speech, Vision, Language, Content Safety, Content Understanding. |",
"reason": "The kind-AIServices-to-Microsoft-Foundry-resource part and projects-as-folders part are grounded, but the terminology mapping shows 'Foundry Tools' is the CURRENT name for the constituent AI services (formerly Azure AI Services), coexisting with the Foundry resource — not the predecessor name the Foundry resource was renamed from — so one load-bearing lineage part is contradicted.",
"file": "skills/ms-ai-security/references/cost-optimization/azure-ai-foundry-cost-governance.md",
"line": 121,
"claim": "The underlying Azure resource (API kind AIServices) is renamed Microsoft Foundry resource (successor/renaming of 'Foundry Tools'); projects function as folders grouping work under one resource",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-security/references/cost-optimization/batch-processing-cost-reduction.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 11,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-security/cost-optimization/batch-processing-cost-reduction.md#2",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/foundry/foundry-models/concepts/deployment-types",
"evidence_quote": "DataZone types: The service processes data only within the Microsoft-specified data zone (US, EU, or Asia Pacific (APAC)). Standard/Regional types: Processed in the deployment region",
"reason": "The two type names Global-Batch and Data Zone Batch are confirmed, but the load-bearing gloss 'regionsbasert' is contradicted: the page's taxonomy explicitly separates data-zone processing (US/EU/APAC zone spanning many regions) from regional processing (single deployment region), so calling Data Zone Batch region-based misstates its data-residency category.",
"file": "skills/ms-ai-security/references/cost-optimization/batch-processing-cost-reduction.md",
"line": 29,
"claim": "To batch deployment-typer: Global-Batch (globalt distribuert) | Data Zone Batch (regionsbasert)",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-security/references/cost-optimization/gpt5-gpt41-pricing-models.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 17,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-security/cost-optimization/gpt5-gpt41-pricing-models.md#1",
"judge_verdict": "not_grounded",
"rule": "",
"evidence_url": "https://learn.microsoft.com/azure/foundry/foundry-models/concepts/models-sold-directly-by-azure?view=foundry-classic",
"evidence_quote": "- 1,047,576 - 300,000 (standard deployments) - 128,000 (provisioned managed and batch deployments)",
"reason": "The page states 300,000 for standard deployments and 128,000 for provisioned managed AND batch deployments, while the claim swaps them (128K for standard/provisioned, 300K for batch).",
"file": "skills/ms-ai-security/references/cost-optimization/gpt5-gpt41-pricing-models.md",
"line": 56,
"claim": "gpt-4.1 kontekstvindu: 1047576 tokens (full), 128000 tokens (standard og provisioned deployments), 300000 tokens (batch deployments)",
"disposition": "outdated"
},
{
"id": "ms-ai-security/cost-optimization/gpt5-gpt41-pricing-models.md#5",
"judge_verdict": "not_grounded",
"rule": "R4",
"evidence_url": "https://learn.microsoft.com/azure/foundry/openai/how-to/reasoning",
"evidence_quote": "Access is no longer restricted for this model.",
"reason": "The live availability table states this for both gpt-5 and gpt-5-codex, so the claimed registration/approval requirement is superseded (the mini/nano/chat 'No access request needed' half still matches, but the load-bearing requirement part is now wrong).",
"file": "skills/ms-ai-security/references/cost-optimization/gpt5-gpt41-pricing-models.md",
"line": 83,
"claim": "gpt-5 og gpt-5-codex krever registrering og godkjenning; gpt-5-mini, gpt-5-nano, gpt-5-chat har ingen registreringskrav",
"disposition": "outdated"
},
{
"id": "ms-ai-security/cost-optimization/gpt5-gpt41-pricing-models.md#9",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/microsoft-copilot-studio/prompt-model-settings",
"evidence_quote": "| Claude Sonnet 4.6 | Standard rate | External model from Anthropic. Context allowed up to 200K tokens. | General |",
"reason": "The current model table lists Claude Sonnet 4.6 (Standard) and Claude Opus 4.6 (Premium) instead of the claimed 4.5 versions and contains no o3 row at all, so multiple load-bearing entries in the claimed rate lineup are superseded or absent.",
"file": "skills/ms-ai-security/references/cost-optimization/gpt5-gpt41-pricing-models.md",
"line": 203,
"claim": "Copilot Credits-takstnivåer: gpt-4.1-mini Basic | gpt-4.1 Standard | gpt-5-chat (preview) Standard | gpt-5-reasoning (preview) Premium | o3 Premium | Claude Sonnet 4.5 (experimental) Standard | Claude Opus 4.5 (experimental) Premium",
"disposition": "outdated"
},
{
"id": "ms-ai-security/cost-optimization/gpt5-gpt41-pricing-models.md#11",
"judge_verdict": "not_grounded",
"rule": "R2",
"evidence_url": "https://learn.microsoft.com/microsoft-copilot-studio/prompt-model-settings",
"evidence_quote": "| Claude Sonnet 4.6 | Standard rate | External model from Anthropic. Context allowed up to 200K tokens. | General | | Claude Opus 4.6 | Premium rate | External model from Anthropic. Context allowed up to 200K tokens. | Deep |",
"reason": "The canonical model enumeration lists Claude Sonnet 4.6 and Claude Opus 4.6 (200K); the claimed Sonnet 4.5/Opus 4.5 are absent, i.e. superseded by 4.6 versions.",
"file": "skills/ms-ai-security/references/cost-optimization/gpt5-gpt41-pricing-models.md",
"line": 217,
"claim": "Claude Sonnet 4.5 og Claude Opus 4.5 er tilgjengelig i Copilot Studio (experimental, 200K kontekstvindu)",
"disposition": "outdated"
},
{
"id": "ms-ai-security/cost-optimization/gpt5-gpt41-pricing-models.md#14",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/azure/foundry/foundry-models/concepts/models-sold-directly-by-azure-region-availability?pivots=standard",
"evidence_quote": "| gpt-5.4 | 2026-03-05 | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ |",
"reason": "The Data Zone Standard Europe table now shows gpt-5.4, gpt-5.5, and gpt-5.6-sol/terra/luna all available in norwayeast, so the load-bearing 'eneste GPT-5 ... er gpt-5.5' part is superseded, even though the no-Regional/PTU parts (gpt-4.1, o-series, GPT-5 family absent in Norway East) still verify.",
"file": "skills/ms-ai-security/references/cost-optimization/gpt5-gpt41-pricing-models.md",
"line": 393,
"claim": "Norway East: kun gpt-4o/gpt-4o-mini tilgjengelig som Norge-resident (Standard/Regional PTU); gpt-4.1, o-serien (o3/o4-mini/o3-mini/o1) og GPT-5-familien finnes ikke som Regional/PTU i Norway East; eneste GPT-5 med EU-residens i Norway East er gpt-5.5 via Data Zone Standard",
"disposition": "outdated"
},
{
"id": "ms-ai-security/cost-optimization/gpt5-gpt41-pricing-models.md#16",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/azure/foundry/openai/how-to/reasoning",
"evidence_quote": "Access is no longer restricted for this model.",
"reason": "The live availability table states this for gpt-5, gpt-5-codex, and gpt-5-pro, contradicting the claimed aka.ms/oai/gpt5access approval requirement and the MCA-E/Default-only restriction for gpt-5-pro; the gpt-5-chat 'Preview (2 versions)' part matches, but the load-bearing access-status parts are superseded.",
"file": "skills/ms-ai-security/references/cost-optimization/gpt5-gpt41-pricing-models.md",
"line": 468,
"claim": "GPT-5-tilgjengelighet: gpt-5 GA (begrenset, krever godkjenning aka.ms/oai/gpt5access) | gpt-5-mini GA | gpt-5-nano GA | gpt-5-chat Preview (2 versjoner) | gpt-5-codex GA (begrenset) | gpt-5-pro GA (begrenset, kun MCA-E/Default-abonnementer)",
"disposition": "outdated"
},
{
"id": "ms-ai-security/cost-optimization/gpt5-gpt41-pricing-models.md#17",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/ai-builder/credit-management",
"evidence_quote": "| Power Apps Premium | 500 | Maximum = 1,000,000 AI Builder credits per tenant. | ... | Power Automate Premium | 5,000 | Maximum = 1,000,000 AI Builder credits per tenant. |",
"reason": "The 500-credit figure applies only to Power Apps Premium; other premium Power Platform plans seed 250 to 5,000 AI Builder credits, and the page adds that these seeded credits are removed on November 1, 2026, so the blanket '500 credits/bruker/mnd i premium-planer' is contradicted even though the gpt-4.1-mini Basic default is correct.",
"file": "skills/ms-ai-security/references/cost-optimization/gpt5-gpt41-pricing-models.md",
"line": 443,
"claim": "AI Builder: prompt builder credits inkludert i premium Power Platform-planer (500 credits/bruker/mnd); default modell gpt-4.1-mini (Basic rate)",
"disposition": "outdated"
}
]
},
{
"file": "skills/ms-ai-security/references/cost-optimization/token-counting-optimization.md",
"batch": "R7.1",
"judged_at": "2026-07-18",
"per_file_verdict": "flagged",
"claim_count": 16,
"verified": null,
"verified_by": null,
"flags": [
{
"id": "ms-ai-security/cost-optimization/token-counting-optimization.md#1",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/dotnet/ai/conceptual/understanding-tokens",
"evidence_quote": "",
"reason": "No learn.microsoft.com page states that o200k_base is the default tiktoken encoding for gpt-4o/o1/o3 (the cited source is GitHub, and two targeted searches found only cl100k_base/BPE mentions on Learn), so the claim can be neither confirmed nor refuted from Microsoft Learn.",
"file": "skills/ms-ai-security/references/cost-optimization/token-counting-optimization.md",
"line": 53,
"claim": "o200k_base er default tiktoken-encoding for gpt-4o, o1, o3",
"disposition": "unsourced"
},
{
"id": "ms-ai-security/cost-optimization/token-counting-optimization.md#6",
"judge_verdict": "source_silent",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/azure/foundry/openai/how-to/prompt-caching",
"evidence_quote": "",
"reason": "The page confirms cache reads are discounted on Standard and 'up to 100% discount on input tokens' for Provisioned, but the load-bearing 50% figure for Standard is never stated on any Learn page - it lives on the JS-rendered Azure pricing page, so it can be neither confirmed nor refuted.",
"file": "skills/ms-ai-security/references/cost-optimization/token-counting-optimization.md",
"line": 148,
"claim": "Cached tokens: 50% rabatt på Standard deployment; opptil 100% rabatt på Provisioned (inkludert i PTU-pris)",
"disposition": "unsourced"
},
{
"id": "ms-ai-security/cost-optimization/token-counting-optimization.md#8",
"judge_verdict": "not_grounded",
"rule": "",
"evidence_url": "https://learn.microsoft.com/en-us/azure/ai-foundry/openai/concepts/models",
"evidence_quote": "| `text-embedding-ada-002` (version 2) | 8,192 | 1,536 | Sep 2021 | ... | `text-embedding-3-large` | 8,192 | 3,072 | Sep 2021 | | `text-embedding-3-small` | 8,192 | 1,536 | Sep 2021 |",
"reason": "The current canonical models table states max request of 8,192 tokens for ada-002 v2, 3-small, and 3-large (the REST reference likewise says '8192 tokens for all embedding models'), which differs from the claimed 8191.",
"file": "skills/ms-ai-security/references/cost-optimization/token-counting-optimization.md",
"line": 263,
"claim": "Token-limit per chunk for embeddings: text-embedding-ada-002 8191 tokens | text-embedding-3-small/large 8191 tokens",
"disposition": "outdated"
},
{
"id": "ms-ai-security/cost-optimization/token-counting-optimization.md#9",
"judge_verdict": "not_grounded",
"rule": "",
"evidence_url": "https://learn.microsoft.com/azure/ai-foundry/openai/concepts/models#fine-tuning-models",
"evidence_quote": "| `gpt-4o-mini` (2024-07-18) | North Central US Sweden Central | - | Input: 128,000 Output: 16,384 Training example context length: 65,536 | Oct 2023 |",
"reason": "The canonical fine-tuning models table states a training example context length of 65,536 tokens for gpt-4o-mini, contradicting the claimed 64,536 (a figure that appears only in the tutorial and is superseded by the spec table); the 128,000 input limit part does hold.",
"file": "skills/ms-ai-security/references/cost-optimization/token-counting-optimization.md",
"line": 327,
"claim": "gpt-4o-mini fine-tuning: training example max 64536 tokens, input limit 128000 tokens",
"disposition": "outdated"
},
{
"id": "ms-ai-security/cost-optimization/token-counting-optimization.md#14",
"judge_verdict": "not_grounded",
"rule": "R8",
"evidence_url": "https://learn.microsoft.com/en-us/azure/ai-foundry/openai/concepts/models",
"evidence_quote": "| `gpt-4.1` (2025-04-14) | ... | - 1,047,576 - 300,000 (standard deployments) - 128,000 (provisioned managed and batch deployments) | 32,768 | May 31, 2024 |",
"reason": "The gpt-4.1 context window is 1,047,576 tokens (~1M) on the live models page, not the claimed 128K (128,000 is only the provisioned/batch deployment cap), so one load-bearing part is wrong even though gpt-5/gpt-5-mini 400K, gpt-4o/gpt-4o-mini 128K, and o3-mini 200K all match.",
"file": "skills/ms-ai-security/references/cost-optimization/token-counting-optimization.md",
"line": 480,
"claim": "Context windows: gpt-5 400K | gpt-5-mini 400K | gpt-4o 128K | gpt-4o-mini 128K | o3-mini 200K | gpt-4.1 128K",
"disposition": "outdated"
},
{
"id": "ms-ai-security/cost-optimization/token-counting-optimization.md#16",
"judge_verdict": "not_grounded",
"rule": "R3",
"evidence_url": "https://learn.microsoft.com/microsoft-copilot-studio/requirements-messages-management",
"evidence_quote": "| Classic answer | 1 Copilot Credit | No charge | | Generative answer | 2 Copilot Credits | No charge | ... Text and generative AI tools (basic) per 10 response 0.1 Copilot Credit per 1K tokens",
"reason": "Copilot Studio bills Generative Answers at a flat per-event rate (2 Copilot Credits, formerly 2 messages), not token-based Azure OpenAI billing; per-1K-token rates apply only to the separate 'Text and generative AI tools' (AI Builder) meters, so the claim's billing frame is replaced by the current credits model.",
"file": "skills/ms-ai-security/references/cost-optimization/token-counting-optimization.md",
"line": 406,
"claim": "Copilot Studio: token-basert billing for Generative Answers (Azure OpenAI), message-basert billing for standard topics; token counting via AI Builder credits",
"disposition": "outdated"
}
]
}
]
}
Reference in a new issue View git blame Copy permalink