From b290ed3ff4d398148be7965dbc5e4ac16579c0ed Mon Sep 17 00:00:00 2001 From: Kjell Tore Guttormsen Date: Wed, 24 Jun 2026 11:38:23 +0200 Subject: [PATCH] =?UTF-8?q?docs(ms-ai-architect):=20KB-refresh=20MEDIUM=20?= =?UTF-8?q?8/33=20=E2=80=94=20regulatory:=20Guardrails-rebranding=20(tidl.?= =?UTF-8?q?=20content=20filters)=20+=20Models=20sold=20by=20Azure-kildetit?= =?UTF-8?q?tel?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .../regulatory-and-compliance-prompting.md | 14 +++++++------- 1 file changed, 7 insertions(+), 7 deletions(-) diff --git a/skills/ms-ai-advisor/references/prompt-engineering/regulatory-and-compliance-prompting.md b/skills/ms-ai-advisor/references/prompt-engineering/regulatory-and-compliance-prompting.md index d83635d..3777eb1 100644 --- a/skills/ms-ai-advisor/references/prompt-engineering/regulatory-and-compliance-prompting.md +++ b/skills/ms-ai-advisor/references/prompt-engineering/regulatory-and-compliance-prompting.md @@ -1,6 +1,6 @@ # Regulatory and Compliance-Aware Prompting -**Last updated:** 2026-02 +**Last updated:** 2026-06-24 **Status:** GA **Category:** Prompt Engineering & LLM Optimization @@ -108,7 +108,7 @@ Du er en helseinformasjons-assistent. HIPAA-regler gjelder: **Azure AI Content Safety Integration:** -Microsoft tilbyr innebygde content filters som evaluerer prompts og completions i sanntid: +Microsoft tilbyr innebygde **Guardrails (tidligere «content filters»)** som evaluerer prompts og completions i sanntid. Default-guardrail for modeller er `Microsoft.DefaultV2`. Kategoriene under er fortsatt gyldige: | Kategori | Beskrivelse | Typisk Threshold (Compliance) | |----------|-------------|-------------------------------| @@ -859,17 +859,17 @@ Før produksjonsdeploy, gjennomfør en 2-dagers compliance sprint: ### Microsoft Official Documentation -1. **Data, privacy, and security for Azure OpenAI** (februar 2026) +1. **Data, privacy, and security for Models sold by Azure in Microsoft Foundry** (juni 2026) https://learn.microsoft.com/en-us/azure/foundry/responsible-ai/openai/data-privacy - → Autoritativ kilde for data processing, abuse monitoring, residency + → Autoritativ kilde for data processing, abuse monitoring, residency. (Sidetittel endret fra «Azure OpenAI» til «Models sold by Azure».) 2. **Govern AI apps and data for regulatory compliance** (februar 2026) https://learn.microsoft.com/en-us/security/security-for-ai/govern → Compliance Manager, Purview integration, EU AI Act readiness -3. **Azure OpenAI Content Filtering** (februar 2026) - https://learn.microsoft.com/en-us/azure/foundry-classic/foundry-models/concepts/content-filter - → Content Safety API, thresholds, custom policies +3. **Guardrails and controls (tidligere «Content Filtering»)** (juni 2026) + https://learn.microsoft.com/en-us/azure/foundry/guardrails/guardrails-overview + → Guardrails/controls, intervensjonspunkter, terskler, `Microsoft.DefaultV2`, custom policies 4. **Azure Data Residency** (februar 2026) https://azure.microsoft.com/explore/global-infrastructure/data-residency