Mucahithan Avcioglu
27 July 2026•Update: 27 July 2026
Nvidia and dozens of global technology and cybersecurity companies on Monday launched the Open Secure AI Alliance to develop and share open-source tools aimed at improving AI safety and security.
The alliance will focus on protecting AI models and autonomous agents by developing technologies for vulnerability detection, identity verification, secure model distribution, software scanning and auditing agent behavior, according to a statement by Nvidia on Monday.
Founding partners include Microsoft, IBM, Cisco, Cloudflare, CrowdStrike, Hugging Face, Red Hat, Salesforce, SAP, Siemens, Dell Technologies, HPE, Palo Alto Networks and the Linux Foundation, among others.
Nvidia said open models and tools allow cybersecurity teams to inspect, adapt and deploy defensive systems on their own infrastructure, reducing reliance on a limited number of closed AI providers and the risk of single points of failure.
The company acknowledged that open AI models could be misused or modified to remove safeguards but argued that similar risks also exist in closed systems.
It said openness should therefore be combined with safeguards, security evaluations, clear restrictions against malicious use and rapid vulnerability remediation.
Nvidia will contribute open models, model weights, datasets and research on AI agent software infrastructure to the alliance.
The company also introduced the open-source Nvidia Labs Object-Oriented Agent, (NOOA) research framework, designed to help developers test, trace, audit and govern AI agent behavior.
Other alliance projects include identity standards for verifying AI agents, secure formats for storing model weights, digitally signed software patches and multi-agent systems capable of detecting and confirming exploitable vulnerabilities in software.
The alliance also urged governments and regulators to treat open AI models and security tools as defensive assets rather than impose broad restrictions on advanced open systems.
It called for increased investment in shared AI security infrastructure, including datasets, evaluation systems, attack simulators and tools used to test models against cyber threats.
