That was the moment Continuous Delivery privilege escalation stopped being a theoretical concern and became a real, urgent problem. Modern CI/CD pipelines are powerful enough to ship code anywhere instantly. They are also powerful enough to grant attackers — or careless insiders — uncontrolled access when security boundaries blur between build, test, and deploy environments.
Privilege escalation in Continuous Delivery happens when a process, tool, or account gains more permissions than it should. In CD pipelines, this can occur through misconfigured build agents, inherited admin roles, exposed secrets, or poor separation between environments. Once a pipeline step or token can touch both test and production, the idea of “safe automation” disappears.
The patterns are well known. Build servers running with elevated permissions. Deployment keys stored in environment variables reused across jobs. Dependencies that download unsigned code during builds. Service accounts that can run arbitrary scripts with production credentials. These weak points often hide in “just works” configurations that teams set up once and never review.
Attackers do not need to breach production directly. They compromise a developer’s workstation, a CI runner, or a shared script. They move from a low-privilege context into the deployment pipeline. The pipeline — trusted by the system — does the rest, pushing malicious code or exfiltrating sensitive data using its own legitimate authorizations.