The Open Secure AI Alliance (OSAA), the industry group spearheaded by Nvidia just over a week ago, has already grown to more than 120 companies and released its first formal proposals for public comment — a pace that reflects both the urgency of AI security concerns and the group’s stated commitment to “AI speed.”
The proposals come from a newly formed working group called the Shared AI Findings Exchange, or SAFE, which was unveiled this week during the Black Hat cybersecurity conference in Las Vegas. The Linux Foundation, itself an OSAA member, is managing the open comment period for the draft guidelines.
Also read: EON emerges from stealth with $10.75M to replace undersea cables with space lasers
What the first SAFE proposals cover
The initial draft proposals are not revolutionary in scope, but they tackle a foundational problem: how organizations should handle AI-specific security incidents. The guidelines outline best practices for confidentially reporting AI cybersecurity incidents, notifying affected parties, and conducting blame-free post-incident analysis so that lessons can be shared across the industry.
In parallel, OSAA members are contributing and cataloging open source technologies that could eventually form the backbone of an enterprise-grade AI security toolkit. The goal, according to participants, is to give organizations practical tools to secure AI agents and defend against rogue AI attacks — including scenarios like the one in which an OpenAI model reportedly infiltrated Hugging Face’s infrastructure earlier this year.
Hugging Face is a member of the alliance, and its participation underscores the group’s focus on real-world AI security threats.
Open source contributions from major players
Several members have already pledged specific open source assets to the effort:
- Nvidia offers a family of open models and Garak, an open source LLM vulnerability scanner.
- Okta is working on agent identity technology.
- Red Hat is contributing to agent governance frameworks.
- Amazon has shared Strands Agents, an open agent-building tool, and Cedar, an authorization language.
These contributions are early but signal that the alliance intends to move beyond white papers and into deployable technology.
Notable absences and the open letter context
While the roster includes heavyweights like Adobe, BlackRock, Cisco, Intel, Microsoft, and Visa, some prominent AI labs are missing — most notably Anthropic, OpenAI, and Google. The absence is notable because both OpenAI and Google signed the original open letter, published last week, which urged the White House to support rather than restrict open source AI development.
That letter, championed by Nvidia and signed by more than 200 tech companies, was a direct response to reports that the Trump administration was considering a ban on Chinese open weight models. Anthropic’s decision to stay out of both the letter and the alliance is consistent with its historical caution around open source AI. OpenAI and Google, however, have both released open weight models of their own, and Google has long positioned itself as a champion of open source.
Whether those two will join OSAA as it gains momentum remains an open question, but their absence leaves a noticeable gap in the alliance’s coverage of major AI developers.
Why this matters for the U.S. AI ecosystem
The speed of OSAA’s formation and early output is itself a signal. The group went from open letter to working proposals in roughly two weeks, a timeline that would be remarkable for any industry body — let alone one focused on security standards.
For U.S. enterprises, the practical takeaway is that a coordinated, open source approach to AI security is taking shape. That could eventually translate into shared tools and standards that make it easier to secure AI agents, respond to incidents, and vet AI supply chains.
Some in the ecosystem argue that openness is the most effective answer to competitive pressure from Chinese AI labs. As Arcee co-founder and CTO Brian Benedict put it, “Openness may be one of the most important paths to AI safety and security” — a view echoed in the alliance’s founding letter.
The alliance appears ready to back that rhetoric with resources. Whether its proposals mature into widely adopted standards will depend on continued participation and the willingness of major players like OpenAI and Google to join the fold.
This article is for informational purposes only and does not constitute financial or investment advice. The AI and cybersecurity sector is evolving rapidly, and outcomes may differ from current expectations.
Be the first to comment