TechCapsules

AI-summarised brief · reviewed before publication

NVIDIA and 36 partner organizations have established the Open Secure AI Alliance to develop open technologies for protecting software and AI agents against cyber threats. The coalition includes major industry players such as Adobe, Cisco, Microsoft, and the Linux Foundation. The alliance aims to build upon the Linux Foundation’s Akrites project and OpenSSF work, providing defenders with inspectable security systems operable on local infrastructure. NVIDIA argues that relying solely on closed AI models creates vulnerabilities, citing a recent Hugging Face incident where closed tools hindered forensic analysis. In contrast, an open-weight model successfully contained the intrusion by analyzing thousands of actions locally. The group asserts that access to both open and closed frontier models is essential for transparency and control. By distributing defensive capabilities across multiple vendors, the alliance seeks to prevent single points of failure and reduce dependency on a few closed providers, ensuring organizations can adapt security tools to their specific needs without sending sensitive data to external providers.