What Is NVIDIA's Open Secure AI Alliance?
The company said the alliance is calling on policymakers to treat open models and security tooling as defensive assets rather than liabilities in AI regulation.
NVIDIA announced on July 27 the formation of the Open Secure AI Alliance, a coalition of enterprise technology companies working to develop and share open tools for artificial intelligence (AI) safety and security. The alliance builds on the Linux Foundation's Akrites initiative and the Open Source Security Foundation (OpenSSF) community.
Founding members include Adobe, Cisco, IBM, Microsoft, Palo Alto Networks, Red Hat, Salesforce, SAP and ServiceNow, alongside cloud providers, cybersecurity vendors and AI research organizations. NVIDIA did not list OpenAI, Anthropic, Google or Meta among the inaugural members.
NVIDIA said the alliance will focus on developing open AI models, evaluation tools and security frameworks that organizations can inspect, modify and deploy within their own environments. The company said open models and harnesses allow security teams to test and verify defensive systems rather than relying solely on closed platforms.
Contributions From Founding Members
NVIDIA said it is contributing open models, model weights, data and agent harness research to the alliance. The company also released the NVIDIA Labs Object-Oriented Agent (NOOA) project on GitHub, a research framework intended to make agent behavior easier to test, trace, audit and govern.
Other members outlined separate contributions. HPE is contributing to SPIFFE/SPIRE, a zero-trust identity framework for verifying AI agents and services. Hugging Face previously contributed its Safetensors model storage format to the PyTorch Foundation in April 2026, months before the alliance formed. Microsoft said its MDASH scanning harness uses multiple AI agents to discover and validate exploitable software bugs. SpaceXAI has open sourced its Grok Build coding agent and plans to release the weights of its Grok model line.
NVIDIA pointed to a recent Hugging Face security incident as a catalyst for the effort. According to Hugging Face's own disclosure, closed AI tools were unable to distinguish attackers from defenders during the incident, so Hugging Face ran the open-weight GLM 5.2 model on its own infrastructure to analyze more than 17,000 actions and contain the intrusion. OpenAI later disclosed that the intrusion originated from its own internal evaluation infrastructure.
The company said the alliance is calling on policymakers to treat open models and security tooling as defensive assets rather than liabilities in AI regulation. NVIDIA said blanket restrictions on open frontier AI systems could concentrate dependence on a small number of closed providers.
NVIDIA said the alliance is inviting additional companies, governments and researchers to contribute to shared infrastructure for AI defense, including datasets, evaluation frameworks and red-teaming tools, as the coalition expands its work.
>
>
Key Takeaways
- NVIDIA launches Open Secure AI Alliance with 37 enterprise partners for AI safety and security.
- Alliance advocates for treating open AI models as defensive assets in regulatory discussions.
- Founding members include major tech firms like Adobe, IBM, and Microsoft, excluding OpenAI and Google.
- NVIDIA contributes open models and research tools to enhance AI testing and security frameworks.
- Members like HPE and Microsoft share innovations to support zero-trust identity and AI bug detection.