The Linux Foundation on Tuesday said it has formed the Agentic AI Foundation (AAIF) to provide vendor-neutral oversight for the development of AI agent infrastructure.<\/p>\n
AI agents are machine learning models empowered to access and manipulate other software, such as web browsers. Despite industry acknowledgment that agents pose security problems<\/a> and IT consultancy Gartner’s insistence that many agent-based enterprise projects will be canceled for lack of business value<\/a>, The Linux Foundation sees value in serving as the Switzerland of free-spending AI firms.<\/p>\n Anthropic, Block, and OpenAI have contributed three projects respectively to this endeavor: Model Context Protocol (MCP)<\/a>, for integrating LLMs with tools; goose<\/a>, an open source AI agent framework; and AGENTS.md<\/a>, the equivalent of a README.md file for machines.<\/p>\n “Bringing these projects together under the AAIF ensures they can grow with the transparency and stability that only open governance provides,” said Jim Zemlin, executive director of the Linux Foundation, in a statement<\/a>. “The Linux Foundation is proud to serve as the neutral home where they will continue to build AI infrastructure the world will rely on.”<\/p>\n Nick Cooper, a member of OpenAI’s technical staff, said in a statement that tools and infrastructure must be trustworthy and accessible for AI agents to reach their full potential.<\/p>\n “OpenAI has long believed that shared, community-driven protocols are essential to a healthy agentic ecosystem, which is why we’ve open sourced key building blocks like the Codex CLI, the Agents SDK, and now AGENTS.md,” he said.<\/p>\n The announcement has been stuffed full of enthusiastic canned quotes lauding the AAIF. The reality of agents has been less worthy of celebration.\u00a0<\/p>\n Microsoft has reportedly<\/a> reduced the growth targets for its Azure Foundry product for building agents, though the company told CNBC<\/a> that “aggregate sales quotas for AI products have not been lowered.”<\/p>\n Google’s Gemini-based coding agent Antigravity was found to be full of security holes<\/a> shortly after its release. It then proceeded to wipe one unfortunate developer’s drive<\/a>. Replit’s AI coding agent managed a similar feat of erasure<\/a> on a production database in July.\u00a0<\/p>\n Such behavior has proven sufficiently concerning that Gartner has called for a corporate ban on agentic browsers<\/a>. And yet companies still show interest in AI-based automation.<\/p>\n Undaunted by past incidents and Air Canada’s humbling chatbot experience<\/a>, Virgin Atlantic on Monday announced<\/a> the availability of its agent-based virtual Concierge. Built with the assistance of AI consulting biz Tomoro, it relies on OpenAI’s Realtime API<\/a> for customer interaction and trip planning.<\/p>\n Other firms like Appdome<\/a>, Coheso<\/a>, EPAM<\/a>, and Lumos<\/a> have recently launched agentic services of their own.\u00a0Sustained by inexplicably optimistic investors, AI agents have taken on a life of their own.<\/p>\n