AI certifications: what changed in 2026

Microsoft replaced its core AI line-up

Microsoft retired the Azure AI Engineer Associate (AI-102) and Azure Data Scientist Associate (DP-100) in June 2026. Neither can be renewed, so every existing credential lapses by mid-2027.

The replacements are split by role: AI-103 for developers building generative AI applications and agents, AI-300 for machine learning operations, and AB-620 for low-code agent building. Azure AI Fundamentals kept its name but moved to a new and more technical exam, AI-901.

AWS shifted towards generative AI

AWS retired the Machine Learning Specialty in March 2026 and made the Generative AI Developer Professional generally available. The ML Engineer Associate is being replaced by a rewritten exam that adds foundation models, retrieval and agent operations under the same certification name.

AWS also made its hands-on microcredentials free, and introduced a business-role certification in beta.

Agent-focused exams arrived

Several providers launched certifications dedicated to agents during the year.

• Google Cloud Professional Agentic Architect, in beta, combining an exam with hands-on labs
• Databricks Certified Context Engineer Associate
• The Linux Foundation’s Model Context Protocol Associate
• IBM watsonx Orchestrate AI Engineer
• Oracle’s four-exam agentic AI track
• Anthropic’s Claude Certified exams, currently limited to partner organisations

Existing exams were rewritten

The Google Cloud Professional Machine Learning Engineer, Databricks Generative AI Engineer Associate, Snowflake’s generative AI specialty and the GitHub Copilot certification all received revised outlines. In most cases the certification name stayed the same.

What this means

For developers, study material from before 2026 is likely to be out of date, so work from the current exam guide. For hiring managers, the date a certification was earned now matters as much as its name.