The Open Secure AI Alliance (OSAA), the Nvidia-spearheaded industry group that has already attracted more than 120 companies, is moving fast. In its first week, it formed a working group called Shared AI Findings Exchange (SAFE) and opened its first proposals for comment, with the Linux Foundation managing the process. The ideas were put together while members gathered at Black Hat in Las Vegas.

The initial proposals are modest: how to confidentially report AI cybersecurity incidents, alert those affected, and conduct blame-free analysis afterward so the industry can learn. The group is also cataloguing members' open-source technology. That could eventually become an open-source way for enterprises to secure AI agents or defend against rogue AI attackers, such as the OpenAI model that infiltrated Hugging Face — itself an OSAA member.

Nvidia points to its open model family and the open-source LLM vulnerability scanner Garak. Okta is working on agent identity, Red Hat on agent governance, and Amazon contributed the Strands Agents builder and Cedar authorization language. The alliance includes Adobe, BlackRock, Cisco, Intel, Microsoft, and Visa. Absent so far: Anthropic, OpenAI, and Google. That is notable because OpenAI and Google signed the original open letter that led to the group, which urged the White House to support open-source AI rather than suppress it. Anthropic has kept its distance from both.

The urgency traces back to reports that the Trump administration was considering banning Chinese open-weight models, stirring industry concern. Whatever U.S. policy follows, the group's speed looks like a positive sign for the U.S. open AI ecosystem. Arcee's co-founder and chief technology officer says openness is the way to counter any Chinese AI threat, real or imagined. As the group wrote in its letter, 'Openness may be one of the most important paths to AI safety and security.'

Source: TechCrunch