On August 4, 2026, the Linux Foundation released the draft for comments of the Shared AI Findings Exchange (SAFE) Guidelines at the Black Hat security conference. Drafted by a working group from the Open Security AI Alliance (with over 120 members), the guidelines were advanced with the participation of companies such as NVIDIA, Cisco, CrowdStrike, Hugging Face, and Red Hat. SAFE aims to transform cybersecurity incidents involving AI agents into shared protection capabilities across the entire ecosystem. By confidentially collecting and analyzing AI security incidents and 'near-misses,' notifying affected parties, identifying recurring issues of security control failures, and issuing evidence-based operational recommendations, it seeks to reduce systemic risks.