What's changed: Added §3.3 "AI Services, Copilot, and Responsible AI" covering the AI area added to AZ-900 in the Jan 2026 update (Azure AI Foundry/OpenAI/AI services/AI Search/Content Safety, the Copilot family, and the six Responsible AI principles)
3.3AI Services, Copilot, and Responsible AI
Understand the Azure AI platform and services (Azure AI Foundry, Azure OpenAI, Azure AI services, AI Search, Content Safety), the Microsoft Copilot family (M365 / Azure / GitHub / Security / Copilot Studio), and the six Responsible AI principles that underpin them. This is the area added to AZ-900 in the January 2026 update.
AI is now a central theme in Azure, and from the January 2026 update it is formally tested on AZ-900. What matters here is not how to build AI but being able to distinguish what services exist and what each is for. Group them into three: the platform and services for building AI, the AI assistants embedded in products (Copilot), and the Responsible AI foundation underneath them all.
3.3.1AI platform and services
Azure AI Foundry is a unified platform to build, evaluate, deploy, and operate AI apps and agents (formerly Azure AI Studio; it includes a playground and model catalog). A flagship service used on it is Azure OpenAI Service, which lets you use large language models (generative AI) such as GPT under Azure’s enterprise security and compliance.
Azure AI services (formerly Cognitive Services) are a family of prebuilt AI capabilities exposed via APIs—vision, language, speech, decision, and more. Because they use pretrained models, you can add image recognition, translation, and the like without training a model yourself. For search, Azure AI Search ingests your own data and retrieves relevant information, forming the basis of search + generative AI (RAG). And Azure AI Content Safety detects and filters harmful text and images, supporting responsible AI.
3.3.2The Microsoft Copilot family
Copilot is the umbrella name for the AI assistants embedded in Microsoft products. There are several, distinguished by "inside which product it helps."
- Microsoft 365 Copilot: assists with drafting, summarizing, and analysis inside M365 apps like Word, Excel, Outlook, and Teams.
- Copilot in Azure: helps manage resources and troubleshoot in the Azure portal/CLI using natural language.
- GitHub Copilot: supports development by completing, generating, and explaining code.
- Security Copilot: assists with security incident investigation and analysis (for Microsoft Security products).
- Copilot Studio (formerly Power Virtual Agents): a low-code tool to build your own copilots/agents.
3.3.3The six Responsible AI principles
Microsoft sets out six principles of Responsible AI to guide how AI is built and used. AZ-900 targets mixing up the meaning of each, so learn them one by one.
- Fairness: treat people equitably, without bias against particular groups.
- Reliability & safety: operate consistently and safely, even in unexpected situations.
- Privacy & security: protect data and handle it securely.
- Inclusiveness: make AI usable by everyone, regardless of ability or background.
- Transparency: make it understandable how the AI reached its decision.
- Accountability: ensure people remain responsible for AI outcomes.
| Service | Role | In a phrase |
|---|---|---|
| Azure AI Foundry | Unified platform for AI apps/agents | Where you build AI |
| Azure OpenAI | Generative AI models (e.g., GPT) for enterprises | Generative AI |
| Azure AI services | Prebuilt AI via APIs (vision/language/speech) | Prebuilt AI parts |
| Azure AI Search | Search over your data + generative AI (RAG) | Search foundation |
| Content Safety | Detect/filter harmful content | Safety filter |
| Copilot | AI assistants embedded in products | AI you use |
Axes for telling them apart: the building side (platform + services) = Azure AI Foundry, OpenAI, AI services, AI Search, the using side (in-product assistants) = the Copilot family, build your own = Copilot Studio. The six Responsible AI principles run through all of them.
Scenario: automating an internal help desk. Ingest internal documents into Azure AI Search, combine with Azure OpenAI generative AI to answer questions (RAG), and filter harmful input/output with Content Safety. Build and evaluate the whole thing on Azure AI Foundry, while staff speed up daily drafting with Microsoft 365 Copilot—combining services whose roles differ by purpose.
Watch the mix-ups: (1) Azure AI Foundry (a platform you build on) vs Copilot (an assistant you use)—"build vs use." (2) Azure OpenAI (generative models) vs Azure AI services (prebuilt API capabilities). (3) Copilot Studio (build your own copilot) vs the various Copilots (ready-made assistants). (4) Mixing up the meaning of fairness / transparency / accountability in Responsible AI.
Q. Azure OpenAI vs Azure AI services? Azure OpenAI provides generative AI like GPT (it generates text, images, etc.). Azure AI services expose prebuilt capabilities via APIs (e.g., image recognition, translation, speech). Think "generate something new = OpenAI / prebuilt recognition or conversion = AI services."
Common points: Azure AI Foundry = unified platform to build AI / Azure OpenAI = generative AI / Azure AI services = prebuilt AI (APIs) / AI Search = search (RAG) / Content Safety = harm detection, Copilot = in-product assistants (M365/Azure/GitHub/Security), Copilot Studio = build your own, and the six Responsible AI principles (fairness, reliability & safety, privacy & security, inclusiveness, transparency, accountability).
3.3.4Section summary
- Build platform + services: Azure AI Foundry (unified base) / Azure OpenAI (generative AI) / Azure AI services (prebuilt APIs) / AI Search (RAG) / Content Safety (safety filter)
- Copilot family: M365 / Azure / GitHub / Security Copilots = in-product assistants. Copilot Studio = build your own
- Six Responsible AI principles: fairness, reliability & safety, privacy & security, inclusiveness, transparency, accountability
Sign in to track progress — Log in.
Quick check
(just a quick review)Q1. Which is the Azure unified platform for building, evaluating, and deploying AI apps and agents?
Q2. Which service provides generative AI models such as GPT under enterprise security and compliance?
Q3. Which is the family of prebuilt AI capabilities (vision, language, speech) callable via APIs?
Q4. Which tool lets you build your own copilots or agents with low code?
Q5. Which Responsible AI principle means making unbiased, equitable decisions across groups?
Q6. Which service detects and filters harmful text and images for generative AI?