On August 12, 2026 (Beijing Time), the U.S. news outlet Axios reported that a consortium of more than 120 organizations, such as NVIDIA, Cisco, and CrowdStrike, has put forward the idea of setting up a standardized safety incident tracking system for AI agents. This system, dubbed the "Shared AI Findings Exchange" (SAFE), was crafted by the Open Security AI Alliance (OSAA), an initiative spearheaded by NVIDIA. Its goal is to tackle industry safety worries stemming from AI agents that have spun out of control. The system welcomes participation from AI model deployers, developers, cloud service and tool providers, independent researchers, critical infrastructure operators, and the like. Government agencies are invited to join as "non-controlling observers." Under this system, members are obligated to report incidents following set timelines. This includes promptly notifying affected organizations, submitting an initial confidential report within four business days, issuing a preliminary factual report within 30 days, and offering updates on fault remediation progress within 90 days. SAFE will collect and scrutinize safety incidents, pinpoint recurring fault types, and encourage the widespread adoption of uniform safety measures throughout the industry.