Vendor-affiliated AI agent tech gets adopted by friendly, funded foundation

The Agentic AI Foundation (AAIF), established in December 2025 by the Linux Foundation to promote vendor-neutral tools for AI agents, has expanded its portfolio by adopting four new projects. These additions bring the total number of projects under the AAIF's protection to ten. The four new adoptees—OpenSandbox, Agent Skills, OpenEnv, and OpenResponses—were donated by Alibaba, Anthropic, Hugging Face, and OpenAI, respectively.

OpenSandbox provides a framework for restricting AI agents' capabilities using client SDKs, tools, protocols, lifecycle hooks, container runtimes, and in-sandbox execution components. Agent Skills offers a standardized text format for guiding AI agents through specific tasks more efficiently, providing necessary context for their objectives.

OpenEnv facilitates the creation of containerized reinforcement learning environments that can be easily scaled and interoperate with various testing demands. OpenResponses presents an open API specification for constructing LLM interfaces compatible with different model providers, standardizing request and response formats, streaming events, and tool invocation patterns.

The AAIF's commitment to keeping these technical specifications open, community-driven, and non-aligned with any specific vendor has attracted significant attention from companies optimistic about the potential of AI agents. Within just a year of its founding, the AAIF has garnered over 300 members, with 60 joining this quarter, including Ant Group, Gates Foundation, MongoDB, Nutanix, Royal Bank of Canada, Vercel, and Vultr as Gold Members.

Although the rapid adoption of AI agents by various industries may appear enthusiastic, it has raised concerns among some sectors, such as US banks, regarding AI agents interacting with customer accounts without proper authorization.