NVIDIA Launches Open Secure AI Alliance to Build Open Tools for AI Security

NVIDIA said Monday that it has launched the Open Secure AI Alliance, a coalition of major technology, cybersecurity and open-source organizations aimed at building and sharing open tools for AI security and incident response. The company framed the effort as a practical response to a growing problem: defenders increasingly need AI security tools they can inspect, control and run inside their own environments rather than rely entirely on closed external services.

The alliance was announced July 27 in an NVIDIA corporate blog post, which said it is intended “to build and share open tools that promote responsible use of and trust in AI.” NVIDIA described the membership as dozens of inaugural partners, including Red Hat, Cloudflare, the Linux Foundation, Microsoft, IBM, CrowdStrike, HPE, Cisco, Hugging Face, Palantir, Salesforce, SAP, Dell Technologies, Databricks, ServiceNow and Snowflake, alongside NVIDIA. The company said the group will focus on open technologies for security, safety and vulnerability response around AI systems.

NVIDIA tied the launch directly to lessons from Hugging Face’s July 16 disclosure of a security incident involving an autonomous agent. In that disclosure, Hugging Face said the agent carried out “many thousands” of actions, and investigators reviewed more than 17,000 recorded events. The company also said it conducted forensic analysis using GLM-5.2, an open-weight model, on its own infrastructure, allowing investigators to analyze attacker-related data without sending sensitive information outside its environment. NVIDIA cited that case as evidence that security teams need open tools they can audit and deploy themselves during an incident.

The new alliance is also being positioned as part of a broader push already underway in open-source security. NVIDIA said the effort builds on the Linux Foundation’s Akrites initiative and OpenSSF community work. The Linux Foundation announced Akrites on June 25, describing it as an effort to defend critical open-source software against AI-enabled cyber threats. That places the Open Secure AI Alliance less as a standalone new structure than as a complementary layer focused on practical tooling and collaboration across AI and security vendors.

In its launch post, NVIDIA said, “The Open Secure AI Alliance — building on the leadership of the Linux Foundation’s Akrites initiative and OpenSSF community work — will work to remediate and disclose vulnerabilities using open technologies.” The company said it plans to contribute open models, model weights, data and new agent harness research, referring to an open-source GitHub project called NVIDIA Labs Object-Oriented Agent, or NOOA. Separate posts published Monday by Red Hat, CrowdStrike and HPE also confirmed their participation in the alliance.

The announcement comes as policymakers debate whether open AI models create new security risks or whether they can also serve as part of the defensive infrastructure needed to respond to attacks. NVIDIA used the launch to stake out a clear position in that debate, arguing that open models, harnesses and security tooling should be treated by policymakers and regulators as defensive assets. It also warned against broad restrictions that, in the company’s view, would leave organizations overly dependent on a small number of closed providers for critical security capabilities.

That policy message gives the launch significance beyond a typical industry partnership announcement. NVIDIA is not only gathering a broad group of recognizable companies around open AI security tools; it is also making the case that openness itself can be a security control, particularly in incident response, where organizations may need to inspect models, verify behavior and keep sensitive evidence inside their own systems. For now, the alliance’s public mission is narrowly framed: share and build open tools that can help defenders respond to vulnerabilities and use AI more safely.