OpenAI co-founds Agentic AI Foundation, donates AGENTS.md
OpenAI
OpenAI just handed over AGENTS.md and co-founded a new Linux Foundation group for agentic AI standards. It's a rare move toward open, shared rules instead of one company owning the agent playbook.
Based on reporting by OpenAI — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
OpenAI is doing something it doesn't do often: giving away control. The company announced it's co-founding the Agentic AI Foundation, a new effort housed under the Linux Foundation, and donating AGENTS.md as its opening contribution. AGENTS.md is the spec that's quietly become a common way for developers to describe how AI agents should behave, what tools they can touch, and what boundaries they shouldn't cross.
The pitch here is interoperability. Right now, every lab building agentic systems tends to invent its own conventions for permissions, tool use, and safety guardrails, which means none of it plays nicely together. By putting AGENTS.md under a neutral foundation rather than keeping it in-house, OpenAI is betting that a shared, vendor-agnostic standard will spread faster and get adopted more widely than anything it could push unilaterally. The Linux Foundation angle matters too — it's the same governance model that turned Kubernetes and dozens of other pieces of critical infrastructure into industry defaults nobody argues about anymore.
And the timing isn't random. Agentic AI has gone from a research curiosity to something enterprises are actually deploying, wiring these systems into codebases, customer service pipelines, and internal tooling. Without common ground rules, that rollout gets messy fast — one company's agent framework won't talk to another's, and safety practices end up reinvented, badly, at every shop. A foundation-backed standard gives smaller players and rivals alike a reason to build on the same foundation rather than a proprietary one owned by a single AI giant.
OpenAI framing this as a safety move is worth sitting with, given how often
My take — AI-written commentary, not fact-checked reporting
Handing AGENTS.md to a neutral foundation is the correct move, and I'll say it plainly: it's smarter than anything OpenAI could gain by hoarding it. Open, boring, well-governed standards are what let an entire ecosystem build safely instead of every vendor reinventing guardrails badly on their own. I just wish more of the industry's safety talk came with this kind of concrete, structural follow-through instead of blog-post promises.
Read more about this at: OpenAI