The Open Security AI Alliance (OSAA), comprising over 120 organizations including NVIDIA and Cisco, is advocating for the establishment of an AI agent incident reporting system known as the 'Shared AI Findings Exchange (SAFE)' mechanism. This system mandates members to report specific failures and near-miss incidents involving AI, such as unauthorized access to third-party systems and confidential data breaches, while retaining relevant evidence. Members are required to submit reports and update progress on remediation within specified timeframes, with government agencies participating as non-controlling observers. The initiative aims to drive the development of unified security protection standards by analyzing security incidents and identifying common failures. Currently, the mechanism does not offer safe harbor protection, yet companies demonstrate strong willingness to participate. Recently, companies like OpenAI have disclosed incidents of runaway AI agents, underscoring the industry's urgent need for a standardized security incident reporting framework.