AI
The Steering Committee’s Role in Kubernetes Governance
Seven people form the overriding governing body of the Kubernetes project. Members serve two-year terms, renewable once, with staggered annual elections so incoming members overlap with outgoing ones. The committee handles broad policy changes, charter updates, and escalations from SIGs. This was the first KubeCon AMA in four or five years because a no-manels policy bars all-male panels, and this is the first time the steering committee has not been all men.
Tightening AI Contribution Policy: What Changes and Why
The committee is hardening its AI policy. Contributors must disclose AI use in PR descriptions. Co-authoring or co-signing commits with an AI tool is banned. Two reasons drive this. First, the Linux Foundation says AI cannot sign a CLA, creating legal liability. Second, “assisted by” and “co-developed with” commit trailers persist in git history and let companies advertise paid tools by pointing at merged PRs to claim Kubernetes endorses them. The ECLA pre-submit automation will block non-compliant PRs automatically.
AI-Generated PRs and the Real Cost to Maintainers
Maintainers receive PRs with code that doesn’t compile, PR descriptions that don’t match their commits, and issue references pointing to hallucinated issue numbers. Commits contain struct members that don’t exist in the codebase. New contributors don’t trigger automatic test runs, so a maintainer must manually approve CI before discovering the PR is worthless. Banning AI co-authors means the ECLA check rejects those PRs before any reviewer wastes time.
SIGs, Working Groups, and How New Ones Form
Kubernetes has no overarching feature roadmap. Individual SIGs own code in specific areas. Horizontal SIGs (docs, security, release) span the whole project. Vertical SIGs own specific code. A new vertical SIG needs an architectural shift large enough to require a new code ownership area. Working groups are designed to close; the batch working group is an exception at three or four years. Several AI-related working groups launched and wound down within the past year. SIG CD is the most recent addition, absorbed because its contributors already overlapped heavily with Kubernetes.
Structural Protections Against Hyperscaler Control
No hyperscaler controls the steering committee. The charter’s maximal representation rule caps any single employer at two of the seven seats. A third election from the same company forces one of the first two to step down. Routine decisions require four votes (a simple majority of seven). Public statements require six (a super majority). The committee aims for all seven to agree before publishing a statement, not just four.
Q&A
What is the hardest and most delightful part of being on steering? Both are the same: taking public credit when the project succeeds and public heat when it doesn’t, acting as a human shield for maintainers and contributors. ▶ 05:00
Are there plans to deprecate Ingress in favor of Gateway API? GA APIs are never removed, only deprecated, and any full deprecation of Ingress would take years of planning with significant backstory that cannot yet be shared. ▶ 21:45
Is the committee open to AI tools that help reviewers? Yes, specifically a pre-review AI linter that catches obvious problems before a human sees the PR, but not AI replacing human reviewers. ▶ 28:18
Notable Quotes
we’re not really negotiable on this one Kat Cosgrove · ▶ 09:48
I’m not going to talk to you Maciej Szulik · ▶ 13:52
Kubernetes by force. That’s not a thing. Kat Cosgrove · ▶ 26:28
Key Takeaways
- The new AI policy bans co-authored commits and “assisted by” trailers due to CLA and marketing rules.
- Maximal representation caps any single employer at two of seven steering seats to block coordinated takeover.
- Working groups are ephemeral by design; new vertical SIGs need a major architectural change to justify creation.
- Maintainers are receiving AI-generated PRs with hallucinated issue numbers and struct members that do not exist.
- The committee supports an AI pre-review linter to triage toil but will not replace human code reviewers.
About the Speaker(s)
Kat Cosgrove is Head of Developer Advocacy at Minimus, focused on growing open source through authentic contribution. Her specialties include approachable introductory content and historical deep dives on DevOps and cloud native technology.
Maciej Szulik is a Staff Platform Engineer at Defense Unicorns with more than 20 years of software development experience. He works on Kubernetes challenges by day and Python side projects by night.
Antonio Ojea is a Software Engineer at Google working on Kubernetes, where he serves as Tech Lead in both SIG Network and SIG Testing. He is one of the project’s top contributors with a track record of leading technical initiatives to improve networking stability.