Open Source Summit + Embedded Linux Conference North America 2026

AI is being discussed as the next evolution of CI/CD, but much of that conversation skips over the realities faced by platform and infrastructure teams responsible for reliability, security, and trust. In practice, introducing AI into pipelines requires restraint, clear boundaries, and a strong understanding of where AI use helps and where it creates risk.
This talk focuses on practical, open-source approaches to using AI in CI/CD pipelines today. We’ll explore patterns where AI acts as a copilot: summarizing pull requests, generating test suggestions, helping engineers interpret CI failures, and enriching security signals while keeping humans firmly in control of decisions.
The session will also cover security concerns, prompt injection risks, secrets exposure, and the importance of treating AI output as untrusted input. We’ll discuss guardrails that help teams experiment safely, such as read-only workflows, explicit review steps, and self-hosted or open-source tooling that avoids sending proprietary code to third-party services.
Attendees will leave with a clear mental model for evaluating AI use cases in their own pipelines and an understanding of the tradeoffs involved.