NVIDIA has launched the Open Secure AI Alliance, bringing together 27 organizations to develop open technologies for detecting, disclosing and fixing cybersecurity vulnerabilities.
Founding members include Microsoft, SpaceX, Dell, HPE, IBM, Red Hat, Hugging Face, the Linux Foundation and NVIDIA. The initiative builds on the Linux Foundation’s Akrites project and the OpenSSF community’s security work.
The alliance argues that researchers need access to both open and closed frontier models to defend against increasingly sophisticated attacks. NVIDIA cited the recent Hugging Face breach involving OpenAI models as an example of the limits imposed by closed systems.
According to NVIDIA, Hugging Face initially tried to use commercial frontier models to investigate and contain the intrusion, but the requests triggered safety restrictions. The company then ran the open-weight GLM 5.2 model on its own infrastructure to analyze more than 17,000 actions and stop the attack.
NVIDIA plans to contribute open models, model weights, datasets and agent frameworks to support the development of new cybersecurity tools. HPE will provide standards and methods for cryptographically verifying AI agents and services, while Hugging Face will contribute its Safetensors format for securely storing model weights.
Microsoft, SpaceXAI, IBM and Red Hat will also provide open-source AI technologies to the initiative.
The alliance is calling on governments and companies to invest jointly in open AI infrastructure. It said regulators should treat open models as defensive assets rather than liabilities, arguing that broad restrictions on open frontier systems could weaken cybersecurity capabilities and concentrate control among a small number of closed providers.
The initiative arrives as the Trump administration reportedly considers broad restrictions on Chinese open-source AI models. OpenAI, Anthropic, Meta and Google are not listed among the alliance’s founding members.
