Nvidia Accelerates Open AI Progress Just One Week After Formation

The Open Secure AI Alliance (OSAA), an industry consortium launched just last week and spearheaded by Nvidia, has rapidly expanded to over 120 companies. This influential group has already established a dedicated working body, aptly named the Shared AI Findings Exchange, or SAFE. This subgroup is actively soliciting public feedback on its initial proposals, with The Linux Foundation, a key member of the alliance, overseeing the management of these submissions.

These foundational guidelines were formulated during the recent Black Hat conference in Las Vegas, a pivotal gathering for the global cybersecurity community, where OSAA members convened.

While the initial proposals may not introduce revolutionary concepts, they lay crucial groundwork. They primarily focus on establishing best practices for the confidential reporting of AI cybersecurity incidents, ensuring timely notification to affected parties, and facilitating a blame-free analytical process to maximize collective learning and improvement across the industry.

Concurrently, OSAA members are actively contributing and cataloging various open-source technologies that could prove beneficial. The long-term vision is for these contributions to coalesce into a robust open-source framework, enabling enterprises to effectively secure their AI agents and defend against malicious AI attacks, such as the incident where an OpenAI model infiltrated Hugging Face. Notably, Hugging Face is also an active participant in this alliance.

Illustrating this collaborative effort, Nvidia has highlighted its extensive suite of open models and its open-source LLM vulnerability scanner, Garak. Okta is developing agent identity technologies, while Red Hat is focused on agent governance. Amazon has contributed both Strands Agents, an open agent building tool, and Cedar, an authorization language. These examples represent just a fraction of the diverse contributions from the alliance members.

The alliance currently boasts an impressive roster of major corporations, including Adobe, BlackRock, Cisco, Intel, Microsoft, and Visa. However, there are conspicuous absences, particularly from prominent AI developers such as Anthropic, OpenAI, and Google.

Intriguingly, both OpenAI and Google were signatories to the original open letter that catalyzed the formation of this group. Published last week, the letter, championed by Nvidia and signed by over 200 tech companies, urged the White House to champion, rather than impede, open-source AI initiatives.

While Anthropic's lack of engagement with both the letter and the alliance to date is not unexpected, it's worth noting that both OpenAI and Google have released their own open-weight models. Google, in particular, is widely recognized as a strong proponent of open source. It remains to be seen whether these major players will join the alliance as it continues to gain momentum.

Operating with remarkable agility, the group's rapid formation comes just weeks after reports surfaced of the Trump Administration considering a ban on Chinese open-weight models, a prospect that sparked considerable industry concern and directly led to the aforementioned open letter.

Regardless of the ultimate outcome concerning Chinese open-weight models in the U.S., the swift and decisive action taken by this influential group appears to be a significant positive development for the U.S. open AI ecosystem. Some within the ecosystem, including the co-founder and chief technology officer of U.S. open-weight AI lab Arcee, contend that such openness is the most effective strategy to counter any perceived or real threats posed by Chinese AI laboratories to the United States.

As the industry group articulated in their foundational letter, "Openness may be one of the most important paths to AI safety and security." The alliance is clearly demonstrating its commitment to these principles by immediately translating its stated intentions into tangible efforts and technological contributions.