Nvidia doesn’t mess around: A week after open AI industry group formed, it’s already showing progress
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Open Secure AI Alliance (OSAA) has grown to over 120 companies in its first week and formed the SAFE (Shared AI Findings Exchange) working group, which is now presenting proposals for open comment with the Linux Foundation managing them.
- OSAA members gathered at the Black Hat conference in Las Vegas this week to draft proposals covering confidential reporting of AI cybersecurity incidents, alerting affected parties, and conducting blame-free analysis of incidents.
- Nvidia has contributed its open model family and the Garak LLM vulnerability scanner, alongside Okta's agent identity tech, Red Hat's agent governance tools, and Amazon's Strands Agents and Cedar authorization language.
- OpenAI, Google, and Anthropic are notably absent from the group, despite OpenAI and Google signing the Nvidia-championed open letter urging the White House to support open-source AI — a letter signed by over 200 companies.
- OSAA members include Adobe, BlackRock, Cisco, Intel, Microsoft, and Visa, with Arcee's CTO arguing that open-source AI is the best counter to perceived threats from Chinese AI labs, the very concern that spurred the Trump administration to consider banning Chinese open weight models.
Why it matters: In one week, 120+ companies built a working group with formal proposals at Black Hat — an unusually fast timeline that gives Nvidia and its allies a first-mover advantage in shaping open-source AI cybersecurity standards. The absence of OpenAI, Google, and Anthropic despite their signatures on the founding letter creates a split between frontier model developers and the rest of the industry, potentially leaving them outside an emerging standard for AI agent security.
Ask SkimNews
