NVIDIA and big tech united for AI security with Open Secure AI Alliance

The decision to focus on open-weight models and frameworks is confirmed in recent incident response episodes, such as the case faced by Hugging Face, demonstrating that the possibility of analyzing and adapting artificial intelligence models on one’s infrastructure represents a critical factor for containing intrusions and conducting timely forensic analyzes.

Founding members include giants such as Adobe, Cisco, Dell Technologies, HPE, IBM, Microsoft, Red Hat, Salesforce, SAP, Siemens and Snowflake, committed to sharing technologies and security standards to avoid centralized points of failure in multi-vendor ecosystems. Each partner makes specific contributions to the defense ecosystemincluding NVIDIA’s NOOA search framework for agent governance, Hugging Face’s Safetensors format for model weight transparency, HPE’s zero-trust identity solutions, Microsoft’s MDASH multi-model scanning systems, and SpaceXAI’s sharing of coding technologies.

Faced with regulatory debates on the management of risks related to open source, The alliance highlights that security cannot depend solely on closed and opaque systems, but requires the mobilization of the entire expert community to test, verify and strengthen collective defensesurging governments and regulators to view open tools as strategic defense assets rather than as drivers of vulnerability.