OpenAI has donated its AGENTS.md specification to the newly formed Agentic AI Foundation, a nonprofit housed under the Linux Foundation.
The company co-founded the foundation alongside other parties and transferred AGENTS.md to its stewardship. AGENTS.md defines how AI agents should interact with software projects — think of it as a README written for automated systems rather than human developers. The foundation's stated mission is to develop open, interoperable standards for agentic AI that prioritize safety. Placing the spec under the Linux Foundation moves it out of OpenAI's direct corporate control, at least on paper.
Donating a specification to a neutral body is a well-worn playbook. Companies do it when they want a standard to spread widely without competitors dismissing it as proprietary lock-in. The Linux Foundation hosts dozens of projects that started at one company and became industry infrastructure — Kubernetes being the clearest example. If AGENTS.md follows the same arc, OpenAI ends up shaping how the entire industry builds agents while appearing to step back.
The skeptical read: open standards donated by the company with the most to gain from their adoption often remain, in practice, that company's standard with a neutral label on the box.