Nvidia said on Aug. 27 it has launched the Open Secure AI Alliance, bringing together nearly 40 technology companies, cybersecurity firms and research organizations to develop open AI security technologies as the industry confronts growing cybersecurity risks tied to artificial intelligence.
The alliance marks Nvidia’s latest effort to expand its role beyond AI chips into AI safety, cybersecurity and industry standards, reinforcing its influence across the broader AI ecosystem.
Members will collaborate on developing and sharing open AI models, model weights, datasets, agent harnesses and security tools designed to identify vulnerabilities, test AI systems and strengthen defenses against emerging threats.
Nvidia also released NOOA, an open-source research project on GitHub that enables developers to test, trace and audit the behavior of AI agents.
The initiative comes as governments and technology companies increasingly debate how to secure rapidly advancing AI systems. Nvidia said open collaboration is essential to discovering vulnerabilities earlier and building stronger defensive capabilities across the industry.
The alliance broadly aligns with Anthropic’s Project Glasswing, another initiative promoting open approaches to AI safety research.
Founding members include Microsoft, IBM, Hewlett Packard Enterprise, Dell Technologies, Palantir Technologies, Salesforce, SAP, Siemens, Cloudflare, CrowdStrike, Databricks, Hugging Face, SpaceX, the Linux Foundation and other technology companies and research organizations.
South Korea’s Naver and SK Telecom also joined as founding members, extending their collaboration with Nvidia on artificial intelligence technologies.
Naver said it will contribute its experience in developing proprietary AI models and participating in open-source AI projects. SK Telecom said it will support the development of AI security technologies as it expands its AI services, AI data-center business and AI agent platform.
The early participation of South Korean companies could position them to help shape emerging global AI security standards and open technology frameworks.
Nvidia also argued that policymakers should treat open AI models, agent harnesses and security tools as defensive assets rather than sources of risk. The company warned that broad restrictions on advanced open AI systems could weaken cybersecurity while concentrating power, dependence and vulnerabilities among a small number of proprietary AI providers.
