Just one week after its formation, the Open Secure AI Alliance (OSAA) - an industry group backed by Nvidia that has already attracted more than 120 member companies - is moving with unusual speed for an organization of its kind. The group has established a working group called the Shared AI Findings Exchange, or SAFE, and is already putting forward formal proposals for public comment. Those proposals are being managed by the Linux Foundation, which is itself a member of the alliance.

The timing of these early developments is deliberate. Many of the group's members convened this week at Black Hat, the prominent cybersecurity conference held annually in Las Vegas, using the gathering as an opportunity to advance the alliance's initial agenda.

Early Proposals Focus on Incident Reporting and Shared Learning

The proposals put forward so far are foundational rather than groundbreaking. They address practical questions around AI cybersecurity: how organizations should confidentially report security incidents involving AI systems, how to notify affected parties in a timely manner, and how to conduct post-incident analysis in a way that avoids assigning blame and instead encourages collective learning across the industry.

While these initial guidelines may not represent dramatic new thinking, they reflect an effort to establish consistent, agreed-upon standards in a space where few formal norms currently exist.

Members Contributing Open Source Tools and Technologies

Beyond the policy proposals, member organizations are also pooling open source technologies that could prove useful in securing AI systems at the enterprise level. The longer-term ambition, as is common with industry coalitions of this type, is that these contributions could eventually form a coherent, open source framework for defending AI agents against threats - including, notably, rogue AI-driven attacks.

One example that has been cited as a reference point is Hugging Face, which was previously targeted by an OpenAI model that managed to infiltrate its systems. Hugging Face is itself a member of the OSAA, giving that example added relevance within the group's discussions.

Several member companies have already made concrete contributions or announced relevant work:

• Nvidia has highlighted its family of open AI models and an open source large language model vulnerability scanner called Garak.
• Okta is developing technology focused on identity management for AI agents.
• Red Hat is working on governance frameworks for AI agents.
• Amazon has contributed two tools: Strands Agents, an open agent-building platform, and Cedar, an authorization language designed for fine-grained access control.

These are only a few examples from what is described as a growing list of contributions across the membership.

A Broad Membership, but Some Notable Gaps

The alliance has attracted a roster of well-known names from across the technology and financial industries. Current members include Adobe, BlackRock, Cisco, Intel, Microsoft, and Visa, among many others. However, three prominent AI companies are conspicuously absent: Anthropic, OpenAI, and Google.

What makes those absences particularly noteworthy is that both OpenAI and Google signed the open letter that directly led to the formation of this group. That letter, published last week, called on the White House to actively support open source AI development rather than restrict it. The letter was championed by Nvidia and signed by more than 200 technology companies.

Anthropic's decision to keep its distance from both the letter and the alliance has not come as a surprise to many observers, given the company's historically cautious stance on open AI development. OpenAI and Google, however, present a more complicated picture. Both companies have released open weight models of their own, and Google in particular has a longstanding reputation as a supporter of open source software. Whether either company will eventually join as the group grows in influence remains to be seen.

Rapid Formation Driven by Policy Concerns

The speed at which all of this has unfolded reflects the urgency that many in the industry feel about the current policy environment. It was only a matter of weeks ago that reports emerged suggesting the Trump administration was weighing a potential ban on Chinese open weight AI models - a prospect that generated significant concern across the technology sector and ultimately prompted the open letter that gave rise to the OSAA.

Regardless of how the question of Chinese open weight models is ultimately resolved in the United States, the swift mobilization of a coalition this large and this well-resourced is being viewed by many as a positive development for the domestic open AI ecosystem. Some voices within that ecosystem have argued that fostering openness and accelerating domestic innovation is the most effective response to any competitive pressure posed by Chinese AI development - whether those concerns are grounded in concrete threats or are more speculative in nature.

The co-founder and chief technology officer of Arcee, a U.S.-based open weight AI lab, has been among those making that argument publicly, suggesting that openness itself represents one of the strongest strategic positions the American AI industry can adopt.

"Openness may be one of the most important paths to AI safety and security," the alliance stated in its founding letter.

Based on the pace of activity in its first week alone, the OSAA appears to be moving quickly to back those words with tangible action - and with the combined technical resources of more than 120 member organizations behind it, the group is positioned to have a meaningful impact on how the industry approaches AI security in the months ahead.