Nvidia-led AI security alliance releases first proposals in its first week

Nvidia's week-old Open Secure AI Alliance (OSAA), now with over 120 members, has released its first proposals for defending against AI agents, managed by The Linux Foundation and presented at the Black Hat conference in Las Vegas. The proposals cover confidential reporting of AI cybersecurity incidents, alerting affected parties, and conducting blame-free analysis. Members are also contributing open source tools, including Nvidia's Garak vulnerability scanner, Okta's agent identity tech, and Amazon's Strands Agents. The group includes Adobe, Microsoft, and Intel, but notably excludes Anthropic, OpenAI, and Google, despite the latter two signing the original open letter urging White House support for open source AI. The fast-moving initiative follows concerns over potential U.S. bans on Chinese open weight models.