In order to improve the security of open source artificial intelligence models and promote their safe applications, NVIDIA joined more than 20 industry giants including Microsoft, Adobe, IBM, Red Hat, SpaceX, OpenClaw and Palantir to officially announce the establishment of the "Open Secure AI Alliance" today.
The alliance aims to build on the Linux Foundation's Akrites project and OpenSSF (Open Source Supply Chain Security Foundation), using open technologies to proactively discover, patch and disclose related security vulnerabilities.
In its statement, Nvidia specifically mentioned the recent cybersecurity incident at Hugging Face. In this incident, closed AI tools were unable to accurately distinguish the behavior of attackers and defenders, thus blocking critical reverse analysis and evidence collection. Faced with the dilemma, Hugging Face deployed and ran the open source weight model GLM 5.2 on its own infrastructure, successfully analyzed more than 17,000 operations, and quickly curbed the spread of the intrusion.
NVIDIA pointed out that this incident revealed a critical technological reality: When network defenders cannot independently review, tune and run cutting-edge AI on their own infrastructure, their emergency response capabilities will be severely tested at critical moments. Therefore, the industry and society urgently need open and controllable cutting-edge defense tools to build a sound security system in a multi-vendor ecosystem and avoid single points of failure.
The core mission of the Open Security AI Alliance is to ensure that cybersecurity defenders around the world have access to and control of trusted open source cutting-edge tools. The alliance emphasizes that the future security ecosystem requires in-depth collaboration between cutting-edge closed models and open source models, which not only provides flexible tool selection, but also ensures transparency, adaptive capabilities, and sovereign control in security-sensitive scenarios.