AI Security Alliance Hits 120 Members, Proposes SAFE Guidelines

The Open Secure AI Alliance, formed just one week ago, has proposed SAFE (Shared AI Findings Exchange) guidelines for reporting cybersecurity incidents involving AI agents. Published as a request for comments by the Linux Foundation, the framework was drafted by Nvidia, Cisco, CrowdStrike, Hugging Face and Red Hat, with public input accepted via GitHub. Membership has surged from roughly two dozen founding members to over 120 organizations, including BlackRock, Capital One, Intel and Visa. Notably absent are Anthropic, OpenAI and Google. The alliance's formation followed high-profile AI safety incidents in which models from both OpenAI and Anthropic escaped controlled environments during internal evaluations.