TL;DR — Key Takeaways
- The Open Secure AI Alliance has moved under the Linux Foundation, giving the NVIDIA-initiated group a more vendor-neutral governance structure.
- OSAA is focused on open technologies for securing AI systems, including models, agents, infrastructure and the software stack around them.
- The move could make it easier for companies to contribute code, research and security practices without concerns about NVIDIA controlling the initiative.
The Open Secure AI Alliance, the NVIDIA-initiated coalition developing open technologies for AI security, has officially transitioned to the Linux Foundation.
An industry consortium for securing AI, the Open Secure AI Alliance (OSAA), certainly sounds like a good idea. But when it’s largely directed by a single company, that’s a little troubling. So NVIDIA decided to make the best move possible and moved the organization to the Linux Foundation.
Since the OSAA was only founded months ago, the only real question is: “Why didn’t they do this in the first place?” As Techstrong’s founder, CEO, and editor-in-chief Alan Shimel observed at the time: “The whole undertaking needs one neutral home.” Amen! And now it has one: The Linux Foundation.
The move places the alliance’s work on open models, security tools, vulnerability disclosure, and AI-agent defenses within the Linux Foundation’s collaborative project structure. It also formalizes a relationship that was present at the alliance’s July launch: the Linux Foundation was an inaugural partner, and OSAA was already building on the Linux Foundation’s Akrites initiative and the Open Source Security Foundation (OpenSSF)’s work.
Looking back, NVIDIA launched OSAA in July with a large group of technology companies, security vendors, model developers, cloud providers, and open-source organizations. The group’s stated aim is to develop and share open tools and techniques to protect AI systems, spanning open-weight models, agent harnesses, and the larger software stack around them.
As Shimel noted at the time, this was “the beginning of an attempt to build a shared security commons for the AI era.” In addition, “The Open Secure AI Alliance appears to understand that AI security is a full-stack responsibility.” Indeed it is. The vendor-neutral OSAA may be the game changer that AI security needs.
Specifically, this change should make it easier for companies to contribute code, specifications, research, and operational experience without framing OSAA as an NVIDIA-controlled program. The Linux Foundation has argued that open models and open weights are central to trustworthy AI. Why? Because they allow users and researchers to inspect, test, audit, modify, and remediate systems rather than relying entirely on black-box services. It has also cast the alliance as part of a broader effort to apply battle-tested, open-source security practices—shared tooling, community review, disclosure processes, and collective defense—to AI systems.
The alliance’s most concrete early project is the proposed Shared AI Findings Exchange (SAFE) working group. This proposed framework would let organizations confidentially learn from AI-related security incidents, near misses, and control failures.
SAFE is intended to transform individual incidents involving AI agents and AI infrastructure into reusable defensive knowledge. Its proposed goals include identifying patterns in failures, alerting affected system owners, improving response practices, and producing evidence-based recommendations without forcing organizations to publicly expose operational details.
In the aftermath of the OpenAI-Hugging Face security fiasco, we now know this kind of approach is essential. By placing it under a vendor-neutral group, everyone will be more likely to help.
In particular, OSAA’s new administration addresses three issues. First, a Foundation-run structure will reassure potential contributors that technical direction, intellectual-property rules, and participation won’t solely be set by NVIDIA.
Second, since AI security requires everyone to cooperate with each other, this is best done under a vendor-neutral forum. Finally, housing the initiative alongside the Linux Foundation’s broader open-source and security ecosystem can give it governance and operational continuity beyond the priorities of an individual sponsor.
The alliance enters the Foundation with a broad initial membership list that includes companies such as Adobe, Amazon, Cisco, Cloudflare, CrowdStrike, GitHub, Google-adjacent open-source communities, Hugging Face, IBM, Intel, Microsoft, Mistral, Mozilla, Palo Alto Networks, Red Hat, Salesforce, SAP, SUSE, and Zscaler, among many others. Indeed, I’m hard-pressed to think of any important AI player who’s not backing the OSAA.
I hope this revamped Foundation can deliver on its promises. This will depend on where it goes from here. For example, we need to see published specifications, working code, interoperable implementations, and a disclosure framework that companies will actually trust with sensitive AI-security information. The SAFE proposal offers an early indication of that direction, but, like everything else, OSAA must move beyond promises to deliver a true, widely supported open AI security framework that can deliver the security goods.