Nvidia-led AI security group launches working group within a week

The Open Secure AI Alliance, spearheaded by Nvidia, came together in response to industry alarm over potential government action against Chinese open weight models. Over 200 companies signed an open letter last week urging the White House to support—rather than restrict—open source AI efforts. Within days, the OSAA was formally established with more than 120 members, including major names like Adobe, BlackRock, Cisco, Intel, Microsoft, and Visa. Notably absent from the roster are Anthropic, OpenAI, and Google, though both OpenAI and Google signed the original open letter that prompted the alliance's formation. The Linux Foundation is managing the alliance's operations.

Within just one week, the OSAA established its first working group, the Shared AI Findings Exchange (SAFE), developed at the Black Hat conference in Las Vegas. SAFE's initial proposals, now open for comment, focus on practical cybersecurity governance: how organizations can confidentially report AI-related security incidents, how to alert those affected, and how to conduct blame-free post-incident analysis so the industry can learn collectively. These guidelines are relatively straightforward rather than technically revolutionary, reflecting a foundational layer of security coordination that the body suggests did not previously exist at scale.

Beyond formal proposals, OSAA members are already contributing open source security tools and technologies. Nvidia is contributing Garak, an open source LLM vulnerability scanner, alongside its broader family of open models. Okta is developing agent identity technology, Red Hat is working on agent governance, and Amazon has contributed both Strands Agents, an open agent building tool, and Cedar, an authorization language. The body suggests this aggregation of member contributions may eventually coalesce into a unified open source framework for enterprises to secure their AI agents and defend against adversarial threats—such as the OpenAI model that the body notes infiltrated Hugging Face, which is also an alliance member.

The alliance's rapid mobilization reflects the urgency driving the effort. The Trump administration's reported consideration of banning Chinese open weight models prompted industry concern just a couple of weeks before the open letter and alliance formation. According to the body, industry voices like the co-founder and CTO of U.S. open weight AI lab Arcee argue that openness and industry coordination represent the best defense against any real or perceived threat posed by Chinese AI labs. The body closes by quoting the OSAA's founding letter: "Openness may be one of the most important paths to AI safety and security," and observes that the group appears ready to put both effort and technology behind that statement immediately.