Note
Access to this page requires authorization. You can try signing in or changing directories.
Access to this page requires authorization. You can try changing directories.
Microsoft 365 enforces change management procedures when both code and non-code changes to its systems are made to maintain its security posture. Any configuration drift from the initial posture can introduce vulnerabilities, break functionality, or disrupt availability. Once an information system within Microsoft 365 is deployed with a robust security posture, detailed change management processes help maintain system integrity.
Many drivers of change exist in Microsoft 365, including new functional or security requirements, feedback from customers, identified vulnerabilities, and audit findings. Regardless of the driver for change, service teams use ticketing or source control tools to document evidence of approval and track all changes.
Source code changes
Engineering and development projects in Microsoft 365 follow Microsoft's Secure Development Lifecycle (SDL) when deploying changes. This software development model includes specific security considerations related to code reviews, tests, and approvals before they're systematically released into the Microsoft 365 environment.

The SDL acts as a framework and includes the identification of possible risks to the finished development project and mitigation strategies that developers can implement and test during the development phases. The framework also includes critical security review and approval checkpoints.
Change identification and planning
Service teams regularly meet to discuss proposed changes, including justification, scope, security impact, priority, dependencies, deployment plans, roles, and responsibilities. They document this information in the change management tracking system. If the team rejects the change, they explicitly document the justification in the ticket for future reference.
Personnel code reviews
Before including new code in a build and deploying it, the code must pass personnel code review. An automated code pipeline attached to each code repository enforces these reviews and can't be circumvented. Once the code receives the required approval, it moves on to the next phase.
Code reviewers check for coding errors, verify that the changes meet the requirements, and perform a security impact analysis. Someone other than the people who developed the code must conduct reviews, enforcing the principle of separation of duties. Preventing the same people from submitting and approving their own code is a critical control that Microsoft strictly enforces. This control greatly reduces the possibility of people single-handedly releasing harmful or buggy code, either intentionally or unintentionally. If reviewers find problems during the code review, they halt the change and have developers resubmit the code with suggested changes and additional testing. Code reviewers can also decide to reject check-in entirely for code that doesn't meet the identified requirements. Once the code is satisfactory, the reviewer provides approval, and the code is checked into the main branch to be included in the next build.
Automated build pipeline and security checks
Once developers commit all changes for the sprint to the main branch, the automated build process begins. The process subjects the code to various automated security checks, including static code analysis, binary analysis, credential and secret scanning, and encryption scanning. Microsoft 365 defines a set of essential tests that each build must pass before deployment to pre-production environments. The process rejects builds that don't pass and sends them back to the development team where adjustments are made until all checks pass. Successful builds proceed to the pre-production environment through an automated deployment pipeline.
Build release
The service team that develops a feature initially releases builds to only themselves. They test stability and functionality before releasing the build to progressively larger test groups in logically isolated cloud environments called rings. After the service team, the build is released to all internal Microsoft 365 groups, followed by release of the build to all internal Microsoft groups. This testing, often referred to internally as dogfooding, allows Microsoft to identify bugs in the true production environment before the build is released to external customers. These testing methods ensure Microsoft's code is secure and functioning as expected before it reaches customers and worldwide deployment. Previous builds are always retained for rollback purposes.
The engineering teams determine the amount of time a build spends in each ring, generally staying for multiple high-load periods before proceeding to the next ring. If all testing is successful in each internal ring, the build is released to customers worldwide, first as a Targeted Release to customer tenants who have opted into that ring, followed by a Worldwide Standard Release.
Non-code changes
Non-code changes are defined as any modifications to Microsoft 365 systems that don't involve creating or editing service source code. This definition includes opening ports, changing Access Control Lists (ACLs), or other changes to the underlying system. In comparison, non-code changes occur less frequently than code changes but still require a high level of scrutiny.

A description of the change is documented along with implementation steps, validation steps, and a rollback plan. Before the change is implemented, at least one person peer reviews the plans for accuracy and security impact. Once approved, the documented plans are implemented. If all validation steps successfully pass, the results are documented in the ticket, and it's marked as resolved.
If the implementation of the change is unsuccessful, the rollback plans are triggered, and the team returns to the planning phase and repeats the process until successful.