diff --git a/skills/ms-ai-security/references/cost-optimization/request-batching-aggregation.md b/skills/ms-ai-security/references/cost-optimization/request-batching-aggregation.md
index df63a71..2cde855 100644
--- a/skills/ms-ai-security/references/cost-optimization/request-batching-aggregation.md
+++ b/skills/ms-ai-security/references/cost-optimization/request-batching-aggregation.md
@@ -1,6 +1,6 @@
 # Request Batching and Response Aggregation
 
-**Last updated:** 2026-05
+**Last updated:** 2026-06
 **Status:** GA
 **Category:** Cost Optimization & FinOps for AI
 
@@ -24,10 +24,11 @@ Denne teknikken er ikke bare en optimaliseringsøvelse — den er ofte nødvendi
 |-----------|-------------|
 | **Input file (JSONL)** | JSON Lines-format med én request per linje, inkludert `custom_id` for korrelasjon |
 | **Global-Batch deployment** | Dedikert deployment-type med 50% lavere pris enn global standard |
+| **Data Zone Batch deployment** | Batch-deployment-type (SKU `DataZoneBatch`) med samme 50% rabatt, men inferens-prosessering begrenset til Microsofts definerte datasone (EU eller US) — relevant for dataresidens |
 | **Enqueued token quota** | Separat quota som ikke forstyrrer online workloads |
 | **24-hour completion window** | Target SLA for batch processing (jobs kan ta lenger, men utløper ikke) |
 | **Output file** | JSONL-resultatfil med responses korrelert via `custom_id` |
-| **Exponential backoff queuing** | Støtte i utvalgte regioner for automatisk retry når token quota er tilgjengelig |
+| **Fail-fast token-quota-håndtering** | I utvalgte regioner feiler nye jobber raskt (fail fast) når enqueued-token-grensen overskrides, slik at klienten programmatisk kan kø-stille og retry-e med exponential backoff |
 
 ### Microsoft Graph JSON Batching
 
@@ -230,14 +231,14 @@ def batch_processor(msgs: List[func.ServiceBusMessage]):
 ### Azure OpenAI Batch API
 
 **Setup:**
-1. Opprett Global-Batch deployment i Azure OpenAI resource
+1. Opprett Global-Batch eller Data Zone Batch deployment i Azure OpenAI resource
 2. Prepare JSONL input file med standard chat completion format + `custom_id`
 3. Upload file via Files API eller Azure Blob Storage
 4. Create batch job med `client.batches.create()`
 5. Poll job status eller subscribe til Event Grid events
 6. Download output file og parse responses
 
-**Regioner med Exponential Backoff Support:**
+**Regioner med fail-fast-kø-støtte (klient-side exponential backoff-retry):**
 - australiaeast, eastus, eastus2, germanywestcentral, italynorth, northcentralus, polandcentral, swedencentral, switzerlandnorth, westus
 
 **Pris:** 50% rabatt vs. global standard (verifiser på Azure pricing page)
@@ -323,10 +324,11 @@ def batch_processor(msgs: List[func.ServiceBusMessage]):
 **Risiko:** Azure OpenAI Batch API kan prosessere data i andre regioner enn der ressursen er hostet.
 
 **Mitigering:**
-1. Bruk European regions (swedencentral, germanywestcentral, switzerlandnorth)
-2. Evaluer TIA (Transfer Impact Assessment) for batch workloads
-3. Vurder Microsoft Foundry med dedikert regional processing (når tilgjengelig)
-4. Alternativt: Azure Machine Learning batch endpoints med regional compute
+1. Bruk **Data Zone Batch** (SKU `DataZoneBatch`): inferens-prosessering begrenses til Microsofts definerte EU-datasone (data prosesseres innenfor EU-medlemsland), mens data-at-rest forblir i ressursens Azure-geografi
+2. Bruk European regions (swedencentral, germanywestcentral, switzerlandnorth)
+3. Evaluer TIA (Transfer Impact Assessment) for batch workloads
+4. Vurder Microsoft Foundry med dedikert regional processing (når tilgjengelig)
+5. Alternativt: Azure Machine Learning batch endpoints med regional compute
 
 ### Budsjettprosesser og Kostnadsforutsigbarhet
 
@@ -356,6 +358,7 @@ def batch_processor(msgs: List[func.ServiceBusMessage]):
 |-----------------|----------------|----------|
 | **Global Standard** | 100% (baseline) | Online chat, real-time inference |
 | **Global Batch** | 50% | Bulk processing, content generation, document analysis |
+| **Data Zone Batch** | 50% | Som Global Batch, men inferens innenfor EU/US-datasone (dataresidens) |
 | **Provisioned Throughput** | Varierer (reservation-based) | Høy throughput, forutsigbar latency |
 
 **Eksempel (GPT-4o, Januar 2026 priser — verifiser på Azure pricing page):**
@@ -493,8 +496,8 @@ Bruk denne matrisen for raskt å avgjøre om batching er riktig:
 ### Microsoft Learn (Verified via MCP)
 
 1. **Azure OpenAI Batch API:**
-   - [Getting started with Azure OpenAI batch deployments](https://learn.microsoft.com/en-us/azure/foundry/openai/how-to/batch) — **Verified 2026-02**
-   - Dekker: JSONL input format, Global-Batch deployment, 50% cost reduction, exponential backoff queuing
+   - [Getting started with Azure OpenAI batch deployments](https://learn.microsoft.com/en-us/azure/foundry/openai/how-to/batch) — **Verified 2026-06**
+   - Dekker: JSONL input format, Global-Batch + Data Zone Batch deployment, 50% cost reduction, fail-fast token-quota-håndtering med klient-side exponential backoff
 
 2. **Microsoft Graph JSON Batching:**
    - [Combine multiple HTTP requests using JSON batching](https://learn.microsoft.com/en-us/graph/json-batching) — **Verified 2026-02**
@@ -512,7 +515,7 @@ Bruk denne matrisen for raskt å avgjøre om batching er riktig:
 
 | Seksjon | Konfidens | Kilde |
 |---------|-----------|-------|
-| Azure OpenAI Batch API | **Verified** | Microsoft Learn MCP (2026-02) |
+| Azure OpenAI Batch API | **Verified** | Microsoft Learn MCP (2026-06) |
 | Microsoft Graph JSON Batching | **Verified** | Microsoft Learn MCP (2026-02) |
 | Azure ML Batch Endpoints | **Verified** | Microsoft Learn MCP (2026-02) |
 | Prismodell (50% rabatt) | **Verified** | Azure OpenAI pricing page (referenced in docs) |