이미지: 마이크로소프트
Why the Age of AI Agents Is Changing Zero Trust
As AI-powered development tools and autonomous agents proliferate, new attack surfaces and trust boundaries are emerging. After first outlining its "Zero Trust for AI" strategy at RSA Conference 2026, Microsoft has now followed up with tools and guidance aimed at moving that strategy into actual implementation. While the earlier announcement laid out architectural direction, this release focuses on providing concrete controls that security, engineering, and platform teams can put to use right away.
Zero Trust Assessment Tool: A New AI Pillar
Previously, the Zero Trust Assessment tool covered the identity, device, network, and data pillars. This update adds three new pillars: AI, Security Operations (SecOps), and Infrastructure. The assessment tool automatically analyzes tenant configurations and activity signals to generate prioritized recommendations, providing both a detailed guide for security teams and an executive summary report.
Assessment results feed directly into the Zero Trust Workshop's "First, Then, Next" framework, designed to turn identified gaps into a phased improvement roadmap. This gives organizations a structured path to establish their current security baseline and identify gaps before adopting AI.
New Workshop Addition: DevSecOps Pillar and AI Memory Guidance
As AI assistants are increasingly used for code generation, package recommendations, and automated infrastructure configuration, concerns have grown that security gaps in development pipelines could lead to greater harm. In response, Microsoft has added a dedicated DevSecOps pillar to its Zero Trust Workshop. This pillar consists of 15 control groups and 91 tasks spanning the full lifecycle from source code to cloud deployment.
The tasks apply the three Zero Trust principles—explicit verification, least privilege, and assume breach—to CI/CD pipelines, source repositories, dependencies, and infrastructure-as-code. Four tasks are directly related to AI-assisted development: code governance, tool allowlist management, data protection, and supply chain security for AI/ML pipelines. The AI pillar also gained new guidance reflecting Microsoft's AI Memory Framework, helping teams manage memory as a security boundary with clear intent, provenance, lifecycle, and user control.
E-book and Pattern/Practice Guide
Microsoft also released an e-book titled "Zero Trust for AI: Rebuilding security controls for autonomous and agentic systems." The book presents a practical framework for applying Zero Trust principles to AI agents, tools, memory, data, and runtime operations, targeting security leaders, architects, and practitioners.
Alongside the e-book, Microsoft released a pattern/practice guide covering five areas: applying least privilege to AI agents, building Zero Trust for source code access, ensuring memory safety in agentic systems, protecting the software supply chain, and governing development security programs. Microsoft said each item was designed to be reusable, much like a software design pattern.
How the Workshop Works and Partner Involvement
The Zero Trust Workshop runs in three stages. First, organizations identify which pillars and stakeholders to assess; then they establish a baseline using the Zero Trust Assessment tool; finally, the workshop turns the findings into a 12- to 24-month roadmap. Partners can use the assessment tool and workshop to deliver customized priority recommendations and phased execution plans for clients, with Microsoft offering partner support through its Frontier Accelerate for Security program.
The source did not disclose pricing or billing details. Specifics such as whether the assessment tool and workshop are free or paid, and which regions are supported, were not included in the information currently available.
