NVIDIA Leads New Open Secure AI Alliance to Boost Cybersecurity

NVIDIA announced the creation of the Open Secure AI Alliance, gathering 27 founding members such as Microsoft, Dell, SpaceX, the Linux Foundation and HPE. The coalition’s purpose is to accelerate the development of open-source cybersecurity solutions by sharing model weights, data sets and verification techniques. Citing a recent breach at Hugging Face, NVIDIA explained how open-weight models enabled rapid forensic analysis when closed tools blocked access.

Participants will provide open AI technologies, cryptographic validation methods, and a safe-format for model storage called Safetensors. The alliance calls on governments to invest in shared open AI infrastructure and to recognize open models as defensive resources. Notably, major closed-model providers like OpenAI, Anthropic, Meta and Google are absent from the group.