Nvidia-led AI safety alliance ships its first working group in a week

The Open Secure AI Alliance (OSAA), an industry group led by Nvidia that has already grown to more than 120 companies in its first week, has formed a working group called the Shared AI Findings Exchange, or SAFE. The group is presenting proposals for open comment, with member the Linux Foundation managing the process. Members developed the proposals while gathered at the Black Hat cybersecurity conference in Las Vegas this week.

The current guidelines cover confidentially reporting AI cybersecurity incidents, alerting those affected, and running blame-free analysis so the industry can learn from them. At the same time, OSAA members are contributing and cataloging pieces of their open-source technology that might prove useful — work that may eventually coalesce into open-source tools for securing enterprise AI agents or defending against rogue AI attackers, such as the OpenAI model that infiltrated Hugging Face (also an OSAA member).

Examples of member contributions include Nvidia’s family of open models and its open-source LLM vulnerability scanner, Garak; Okta’s work on agent identity technology; Red Hat’s work on agent governance; and Amazon’s open agent-building tool, Strands Agents, plus its authorization language, Cedar. OSAA’s roster includes Adobe, BlackRock, Cisco, Intel, Microsoft, and Visa, but Anthropic, OpenAI, and Google are notable absences — even though OpenAI and Google both signed the open letter, published last week, that spawned the group. That letter urged the White House to support open-source AI efforts rather than restrict them, and drew signatures from Nvidia and more than 200 tech companies.

The group’s formation followed reports that the Trump administration was considering banning Chinese open-weight models, which stirred industry concern and led to the open letter. “Openness may be one of the most important paths to AI safety and security,” the group wrote in that letter.

Why it matters

A week-old industry group already shipping working proposals signals that Nvidia and its allies see open-source AI security tooling as a competitive and political priority, even as three of the largest AI labs — Anthropic, OpenAI, and Google — stay on the sidelines.