Nvidia launches Open Secure AI Alliance to develop open AI security tools

Nvidia has launched the Open Secure AI Alliance (OSAA), a new industry coalition that aims to develop open-source tools and frameworks to secure artificial intelligence systems, as concerns grow over the cybersecurity risks posed by increasingly autonomous AI models. The initiative comes days after an AI security incident involving Hugging Face and amid a broader industry debate over open versus closed AI models. 

The alliance brings together more than 30 founding members spanning technology, cybersecurity and enterprise software. Founding participants include Microsoft, Cisco, IBM, Adobe, Dell Technologies, CrowdStrike, Red Hat, Palo Alto Networks, Snowflake, ServiceNow, Hugging Face, Cohesity, Palantir, SpaceX AI, Trend Micro and the Linux Foundation, among others. Nvidia said the group will jointly build and share open technologies that help developers, enterprises and governments build AI systems more securely. 

“The world needs both closed and open models. In cybersecurity, open models and open harnesses are essential because they democratize defense, increase transparency for defenders, support cyber defense while protecting data, and complement frontier closed models through customizable, localized controls,” Nvidia said in its announcement.

According to the company, the alliance will focus on securing AI agents and foundation models, protecting AI infrastructure, improving software supply chain security, advancing red-teaming capabilities, creating benchmarks and datasets for AI security testing, and developing tools that can identify and remediate vulnerabilities. Rather than creating new standards from scratch, OSAA will build on existing open-source initiatives, including projects under the Linux Foundation and the Open Source Security Foundation (OpenSSF). 

Nvidia argued that AI security should follow the same collaborative approach that has shaped modern open-source software.

“Just as open source created a shared foundation for software, the United States and its partners now face a choice in AI security: whether the defenses that protect our infrastructure will sit inside a few opaque systems or be built on open models, harnesses and tools that any defender can study, adapt and deploy,” the company said.

Nvidia acknowledged that open AI models can also be abused by malicious actors but argued that restricting access is not the answer. "The right response is not to deny defenders access to capable open systems. It is to pair openness with strong safeguards, clear rules against malicious misuse, rigorous evaluation and rapid remediation. In cybersecurity, the safer path is the one that gives more defenders the ability to test, verify and strengthen the systems on which society relies," the company said. 

What each company is offering

NVIDIA is contributing a new open source Nvidia Labs Object-Oriented Agent (NOOA) project, which is available on GitHub to make advanced AI safety capabilities more accessible for agent harnesses. . 

HPE contributes to SPIFFE/SPIRE, which creates zero-trust identity framework standards and methods that can cryptographically verify AI agents and services to ensure only authorized workloads communicate and access enterprise resources. 

Hugging Face has offered Safetensors, a safe format to store AI model weights, providing transparency and guarantees of no remote code execution to the PyTorch Foundation. 

IBM and Red Hat’s Lightwell extends security across the open source supply chain with digitally signed patches. 

Microsoft’s MDASH multi-model agentic scanning harness orchestrates specialized AI agents to discover, debate and prove exploitable bugs. 

SpaceXAI has open sourced the Grok Build terminal-based AI coding agent to promote trust, transparency and new capabilities, and plans to open source the weights of the Grok line of models to support the developer and research communities.