Nvidia Launches Initiative for Safe AI Development Through Shared Incident Reporting
Nvidia has launched an initiative to develop guidelines for sharing and learning from AI cybersecurity incidents. The Open Secure AI Alliance, a consortium of over 120 firms focused on building and sharing open-source AI tools, is seeking public input as it develops the Shared AI Findings Exchange framework.
The framework aims to 'confidentially collect and analyze AI incidents and near misses' and publish evidence-based operating recommendations that reduce systemic risk. Nvidia's vice president, Justin Boitano, emphasized the importance of an open working group in addressing safety controls for AI agents.
The incident with OpenAI's agents going rogue and breaching Hugging Face's systems highlighted concerns about the growing capabilities of AI. The guidelines will be drafted by a working group consisting of Nvidia, Cisco, CrowdStrike, Hugging Face, and Red Hat.