Nvidia has formed the Open Secure AI Alliance, a coalition to build open security tooling for AI systems, announced on July 27 in the days after an OpenAI agent breached Hugging Face’s production infrastructure. OpenAI, Google, Anthropic and Meta are absent from the membership.
Coverage at launch put the founding roster at between roughly 30 and 40 organizations, with one count at 37. Nvidia’s partner list has since grown past 60. Named participants include Microsoft, IBM, Red Hat, Cisco, CrowdStrike, Palo Alto Networks, Cloudflare, Databricks, Palantir, SAP, Siemens, Hugging Face, Mistral, Perplexity, Thinking Machines Lab, Nous Research and the Linux Foundation. The stated purpose is to develop and share open technologies, techniques and tools to safeguard software and agents.
The alliance’s focus follows a failure exposed during the breach response. When Hugging Face responders tried to analyze the intrusion using frontier models behind commercial APIs, provider guardrails blocked the requests, because the work required submitting live exploit payloads and command-and-control artifacts that the models could not distinguish from an attacker’s own. Hugging Face completed the forensic reconstruction using GLM 5.2, an open-weight model from Chinese developer Z.ai, run on its own hardware. That approach kept attacker data inside the company and reconstructed a timeline from more than 17,000 recorded events in hours.
Z.ai released GLM 5.2 as an open-weight model and trains its GLM series on a gigawatt-scale data center built entirely on Chinese chips.
Nvidia is contributing open models, weights and data, along with NOOA, or NVIDIA Object-Oriented Agents, an Apache 2.0 research framework published on GitHub under the NVIDIA-NeMo organization. It represents an agent as a Python object, with methods as available actions and docstrings as model instructions, and is intended to make agent behavior easier to test, trace, audit and govern.
The launch materials did not include a charter, a governing board, technical workstreams, a delivery schedule or a shared repository. The Linux Foundation is described as an inaugural partner providing neutral ground for competing organizations, and Nvidia did not state that the alliance is formally hosted or governed as a Linux Foundation project.
The absent companies have taken varying positions on open weights. Nvidia, Microsoft, Meta, Google and OpenAI all signed a separate industry letter on July 24 arguing that downloadable models benefit defenders and reduce vendor dependence. Anthropic did not sign that letter and has not joined the alliance.
The coalition adds to a set of competing efforts to set security standards for autonomous systems, including the FIDO Alliance’s work on agentic authentication. Nvidia has said the alliance will work to remediate and disclose vulnerabilities using open technologies. NOOA is available on GitHub now.
Sources: NVIDIA, The Hacker News, CNBC
–
By the Control Plane Editorial Team