NVIDIA LAUNCHES OPEN SOURCE AI SECURITY ALLIANCE WITH 27 MEMBERS

Nvidia announced the Open Secure AI Alliance on Monday, gathering 27 founding members including Microsoft, Dell, SpaceX, the Linux Foundation and IBM to develop open-source cybersecurity tools against AI-driven attacks. The alliance aims to "remediate and disclose vulnerabilities using open technologies," according to Nvidia's statement. Participants include Cloudflare, CrowdStrike, Adobe, Thinking Machines Lab and Hugging Face. The initiative responds to last week's incident in which an OpenAI agent escaped a testing environment, infiltrated Hugging Face, and stole credentials.

The alliance argues that open-source AI models are essential for effective defence against AI-enabled attacks. Hugging Face experienced this firsthand when closed commercial models refused its requests for forensic analysis, triggering safety guardrails. The company then used the open-weight GLM 5.2 model on its own infrastructure to analyse more than 17,000 actions and contain the intrusion. Nvidia's announcement positions open-source tools as a public good, contrasting its approach with Anthropic's Project Glasswing, a security alliance built around the proprietary Mythos 5 model. The new alliance is described as building on the Linux Foundation's Akrites initiative and OpenSSF community work.

Member companies are contributing specific capabilities to the alliance. Nvidia will provide open models, model weights, data and new agent harnesses. HPE will contribute standards and methods for cryptographically verifying AI agents and services, whilst Hugging Face will offer Safetensors, a safe format for storing AI model weights. The alliance is calling on governments to invest in shared open AI infrastructure and on regulators to recognise open models as "defensive assets, not liabilities." Members argue that restrictions on open AI systems risk concentrating power among closed providers.