The Open Security AI Alliance (OSAA), established just a week ago by NVIDIA, is rapidly advancing the AI security ecosystem. The alliance has already grown to more than 120 companies and announced the establishment of the "Shared AI Findings Exchange" (SAFE) working group, which has started collecting proposals for handling AI security incidents. The Linux Foundation is responsible for managing this initiative.
During this week's Black Hat conference, the alliance further refined its plans, focusing on aspects such as confidential reporting of AI cybersecurity incidents, notification mechanisms for affected parties, and blameless post-mortem analysis. These efforts aim to promote the sharing of security experiences and best practices within the industry.
In addition to developing collaboration standards, OSAA members continue to contribute open-source security capabilities, with plans to gradually form an open-source AI security toolchain available for enterprise deployment. This will help protect AI agents and defend against malicious AI attacks. NVIDIA provides open models and the open-source LLM vulnerability scanning tool Garak. Okta advances Agent identity technology, Red Hat develops Agent governance solutions, and Amazon contributes projects such as Strands Agents and the authorization language Cedar.
The alliance members include companies such as Adobe, BlackRock, Cisco, Intel, Microsoft, and Visa. However, OpenAI, Anthropic, and Google have not yet joined. Despite this, both OpenAI and Google signed an earlier open letter initiated by NVIDIA, supported by over 200 tech companies, advocating for open-source AI.
This rapid formation of the alliance is closely related to recent discussions in the U.S. about open-source AI regulatory policies. Industry experts believe that the OSAA's accelerated progress in promoting open collaboration and security standards could further strengthen the U.S. open-source AI ecosystem and enhance AI security capabilities and industrial competitiveness through open sharing.
