Linux Foundation Launches Agentic AI Foundation for Open Standards

The Linux Foundation announced the formation of the Agentic AI Foundation (AAIF) on December 9, 2025, to establish a neutral, open framework for the evolution of agentic AI. The organization provides legal and technical support to ensure transparency and stability through open governance, aiming to foster interoperability between different AI providers and systems.

Founding members include Amazon Web Services, Anthropic, Bloomberg, Cloudflare, Google, IBM, JetBrains, Microsoft, OpenAI, and Salesforce. To support these shared standards, several companies donated key technologies to the AAIF. Anthropic transferred ownership of its Model Context Protocol (MCP), OpenAI donated the AGENTS.md open format for guiding coding agents, and Block provided its Goose framework. Google also donated its Agent2Agent protocol to the foundation.

This initiative follows a year of rapid transition in 2025, where AI agents moved from theoretical research to functional infrastructure. While the industry saw the rise of open-weight models like DeepSeek-R1 and the development of agentic browsers, the shift also introduced security risks. These included vulnerabilities to indirect prompt injections and the misuse of Anthropic's Claude Code in a cyberattack. The AAIF now seeks to govern these shared standards and best practices to facilitate the integration of AI agents across business operations and the broader economy.