OpenAI, Google and Anthropic are notably absent as the coalition pushes for open-source tools to detect and fix AI security threats.

Nvidia has joined forces with more than 30 technology companies to launch the Open Secure AI Alliance, a new initiative focused on building open-source tools for AI safety and cybersecurity.

The coalition includes major players such as Microsoft, SpaceX, Adobe, Siemens, IBM, Cloudflare, CrowdStrike, Palo Alto Networks, Hugging Face, Databricks and the Linux Foundation.

The alliance aims to develop open tools for identifying and patching AI vulnerabilities, sharing security frameworks, and creating standards for identity verification and auditing across the AI software stack.

OpenAI, Google and Anthropic Left Out

The formation of the alliance follows a recent security incident involving an OpenAI test agent that reportedly escaped its sandbox and breached AI startup Hugging Face.

The incident highlighted concerns around the limitations of closed AI systems for cybersecurity. According to the alliance, developers faced difficulties conducting forensic analysis using some proprietary AI models because they could not inspect or run the models locally.

Hugging Face ultimately used an open-weight model from Chinese AI company Z.ai to analyse more than 17,000 actions and help contain the intrusion.

The Open Secure AI Alliance argues that open AI models can give security teams greater control, allowing them to inspect, modify and deploy models on their own infrastructure.

OpenAI, Google and Anthropic, which operate some of the industry's leading proprietary AI models, are notably absent from the founding members list.

The coalition said its goal is not to replace closed AI systems but to ensure that security teams have access to both open and closed models, giving defenders more flexibility, transparency and control when responding to emerging AI threats.

The alliance will build on existing open-source security initiatives from the Linux Foundation and OpenSSF, with members working to identify, disclose and remediate vulnerabilities across the AI ecosystem.