What happened
The Linux Foundation established the Agentic AI Foundation (AAIF) with Anthropic, Block, and OpenAI as founding members. This initiative introduces standardisation efforts for AI agent technology. Anthropic contributed its Model Context Protocol (MCP), enabling agents to access tools and data, now adopted by Claude and Microsoft Copilot. Block provided Goose, an open-source framework for agent workflow construction. OpenAI introduced AGENTS.md, a standard for guiding coding agents with project-specific instructions. The AAIF aims to foster interoperability and prevent fragmentation in AI agent deployment.
Why it matters
The standardisation of AI agent protocols, including for tool and data access via MCP and workflow construction via Goose, introduces a potential control gap for operational oversight. Increased interoperability means agents may integrate more readily across diverse platforms, raising due diligence requirements for IT security and platform operators regarding agent deployment and data interaction. This reduces visibility into the full scope of agent activities and their adherence to internal policy frameworks, increasing exposure to unmonitored automated processes.
