Nvidia-led AI security group OSAA unveils first proposals at Black Hat

The Open Secure AI Alliance (OSAA), an initiative launched by Nvidia just a week ago, has swiftly produced its initial proposals for public feedback. This coalition has expanded to encompass more than 120 companies and unveiled a working group named the Shared AI Findings Exchange (SAFE) at the Black Hat conference in Las Vegas. The proposals prioritize confidential reporting of AI-related cybersecurity incidents, notifying impacted parties, and performing blame-free analyses to enhance learning. OSAA members are also sharing open-source technologies, including Nvidia's open models, Garak's vulnerability scanner, Okta's identity technology, Red Hat's governance tools, and Amazon's Strands Agents and Cedar authorization language. Key participants include Adobe, BlackRock, Cisco, Intel, Microsoft, and Visa, while Anthropic, OpenAI, and Google have opted not to join, despite the latter two signing the original letter that initiated the group. This letter, released last week, called on the White House to back open-source AI initiatives and garnered signatures from over 200 tech firms. The group's prompt actions coincide with the Trump administration's discussions about potentially banning Chinese open weight models, raising industry alarms. The Linux Foundation is overseeing the proposals, and Arcee, co-founder and CTO of the U.S. open weight AI lab, stressed the importance of openness for AI safety and security.