The Linux Foundation has released a draft for public comment on the 'Shared AI Discovery Exchange Guidelines'.

According to NVIDIA’s blog, as the annual Black Hat Security Conference kicks off today in Las Vegas, members of the Open Secure AI Alliance (OSAA) are developing new guidelines to strengthen cybersecurity protections for agent AI. The alliance now has over 120 institutional members. The Linux Foundation today released a draft for comment of the Shared AI Discovery and Exchange (SAFE) Guide, a set of proposed guidelines aimed at turning cybersecurity incidents in the agent AI space into shared protective capabilities for the entire ecosystem. The SAFE Guide was drafted by the Open Secure AI Alliance working group. Alliance members including NVIDIA, Cisco Systems, CrowdStrike, Hugging Face, and Red Hat are collaborating with the Linux Foundation to support the initial proposal. The SAFE Guide includes several recommended measures: confidentially collecting and analyzing AI security incidents and "near-miss events"; notifying affected stakeholders; identifying recurring failures in security controls; and publishing evidence-based operational recommendations to reduce risks across the entire system.