Nvidia Creates Global AI Security Network, But OpenAI and Anthropic Are Missing

Nvidia and 36 other companies believe cybersecurity teams need access to AI systems they can independently run, inspect, and adapt. Major AI firms such as OpenAI, Anthropic, and Google are not included in the newly formed alliance.

U.S. chipmaker Nvidia and a coalition of 36 technology companies launched the Open Secure AI Alliance on Monday to develop open-source security tools for artificial intelligence. The initiative was created after a recent incident highlighted concerns that closed AI models could prevent organizations from fully investigating attacks against their own systems.

The alliance brings together companies and organizations including Microsoft, IBM, Red Hat, Cloudflare, CrowdStrike, Palantir, Databricks, Hugging Face, SpaceXAI, and the Linux Foundation. It builds on existing Linux Foundation efforts, including the Akrites initiative and OpenSSF projects. However, OpenAI, Anthropic, and Google — developers of some of the most advanced proprietary AI models — are absent from the founding group.

The alliance was partly formed in response to the recent Hugging Face security incident.

OpenAI said AI models involved in an internal cybersecurity benchmark, where safety restrictions were intentionally reduced for testing, broke out of a controlled environment and gained the ability to execute commands on Hugging Face’s production infrastructure.

According to Nvidia, the recovery process encountered another challenge because closed AI tools were unable to properly differentiate between malicious activity and legitimate defensive investigations. Hugging Face instead used GLM 5.2, an open-weight model created by Chinese AI developer Z.ai and deployed on its own infrastructure, to analyze more than 17,000 actions and help contain the intrusion.

Nvidia said defenders need the ability to access, customize, and operate advanced AI models directly on their own infrastructure, particularly during security emergencies where rapid response is crucial.

Alliance members are contributing various security technologies. Nvidia released NOOA, a framework designed to evaluate and audit AI agent behavior, through GitHub. Microsoft contributed MDASH, a platform that uses multiple AI agents to identify exploitable weaknesses, while SpaceXAI open-sourced its Grok Build coding agent and said it intends to release Grok model weights.

The alliance launches as cybersecurity threats continue to rise globally, with crypto networks and digital wallets remaining popular targets because successful attacks can generate significant financial gains and blockchain transactions are typically irreversible.

Last week, four protocols lost more than $35 million combined in attacks, including AFX, Verus, and Bitcoin scaling project B². None of the incidents involved defeating cryptographic protections. Instead, attackers exploited trusted permissions and control systems — the same type of complex, multi-step activity where AI agents are becoming increasingly effective.

Unlike traditional company breaches, funds stolen from compromised blockchain contracts usually cannot be recovered after transactions are confirmed.