docs(architect): weekly KB update — 66 files refreshed (2026-04)

Updated 66 stale knowledge base reference files (10 critical, 56 high)
across all 5 skills using Microsoft Learn MCP research.

Key factual updates:
- Groundedness Detection API: `correction` → `mitigating` param,
  `correctedText` → `correctionText` (breaking change)
- Copilot Studio: GPT-4.1 mini now default (was GPT-4o mini);
  Claude Sonnet 4.5 + Opus 4.5 added (experimental, 200K ctx)
- Agentic Retrieval: still public preview; 50M free tokens/month
- Azure security baselines: "Cognitive Services" → "Foundry Tools"
- Databricks: Delta Live Tables → Lakeflow Spark Declarative Pipelines
- MLflow 3 GenAI: new Feedback/Expectation data model
- Token tracking doc: "Azure OpenAI in Foundry Models through a gateway"
- Agent Registry: Risks column (M365 E7), Graph API (preview)
- Copilot DLP: new Entra AI Admin + Purview Data Security AI Admin roles
- ISO/IEC 42001: scope expanded to M365 Copilot, Foundry, Security Copilot
- Zero Trust: CAE now via Conditional Access, Strict Location Enforcement
- Purview: new Fabric Copilots/agents governance section
- AG-UI HITL: ApprovalRequiredAIFunction (C#), @tool approval_mode (Python)

All files: Last updated → 2026-04, *(Verified MCP 2026-04)* markers added.
Build registry: 1341 URLs from 387 files (+2 new URLs).

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

skills/ms-ai-security/references/ai-security-engineering/jailbreak-prevention-production.md

@@ -1,6 +1,6 @@
 # Jailbreak Prevention in Production
 
-**Last updated:** 2026-02
+**Last updated:** 2026-04
 **Status:** GA
 **Category:** AI Security Engineering
 
@@ -300,23 +300,50 @@ curl --location '<endpoint>/contentsafety/text:analyzeCustomCategory?api-version
 }'
 ```
 
-### API Management Integration
+### API Management Integration *(Verified MCP 2026-04)*
 
-**llm-content-safety policy** for LLM requests:
+**llm-content-safety policy** for LLM requests — nå med nye attributter:
 
 ```xml
 <policies>
     <inbound>
-        <llm-content-safety backend-id="content-safety-backend" shield-prompt="true">
+        <!-- Sjekk requests OG responses (enforce-on-completions) -->
+        <llm-content-safety backend-id="content-safety-backend"
+                            shield-prompt="true"
+                            enforce-on-completions="true">
             <categories output-type="EightSeverityLevels">
                 <category name="Hate" threshold="4" />
                 <category name="Violence" threshold="4" />
+                <category name="SelfHarm" threshold="4" />
+                <category name="Sexual" threshold="6" />
             </categories>
+            <!-- Egendefinerte blokkeringslister -->
+            <blocklists>
+                <id>company-jailbreak-blocklist</id>
+            </blocklists>
         </llm-content-safety>
     </inbound>
+    <!-- Alternativt: sett i outbound for å sjekke LLM-svar -->
+    <outbound>
+        <llm-content-safety backend-id="content-safety-backend"
+                            window-size="8000"
+                            window-overlap-size="200">
+            <categories output-type="EightSeverityLevels">
+                <category name="Hate" threshold="4" />
+            </categories>
+        </llm-content-safety>
+    </outbound>
 </policies>
 ```
 
+**Nye attributter (2026-04):**
+- `enforce-on-completions="true"`: I inbound-seksjonen — validerer også LLM-responser
+- `window-size`: Tegnvindusstørrelse for responssjekk (default 10 000)
+- `window-overlap-size`: Overlapp mellom vinduer (for lange responser)
+- `<blocklists>`: Legg til Content Safety-blokkeringslister direkte i policyen
+- Støttede kategorier: `Hate`, `SelfHarm`, `Sexual`, `Violence`
+- Policyen kan brukes i `inbound` og `outbound`, og kan defineres **flere ganger** i samme policy definition *(Verified MCP 2026-04)*
+
 ## Produksjonsovervåking
 
 ### Metrics to Track
@@ -527,11 +554,11 @@ print(f"Jailbreak resistance score: {results['jailbreak_resistance']}")
 
 ### Verification Status
 
-- ✅ **All Microsoft Learn URLs verified**: 2026-02
+- ✅ **All Microsoft Learn URLs verified**: 2026-04 (re-verifisert via MCP)
 - ✅ **API examples tested**: Azure OpenAI API version 2024-09-01
 - ✅ **Production deployment patterns**: Based on Microsoft AI Playbook
 - ✅ **Norwegian public sector alignment**: Cross-referenced with Utredningsinstruksen, NSM, Digdir guidelines
 
 ### Research Date
 
-Denne referansen er basert på Microsoft Learn-dokumentasjon hentet **2026-02-05** via `microsoft-learn` MCP server (6 searches, 3 full document fetches).
+Denne referansen er basert på Microsoft Learn-dokumentasjon hentet **2026-02-05** og re-verifisert **2026-04** via `microsoft-learn` MCP server. *(Verified MCP 2026-04)*