Nvidia leads Open Secure AI Alliance for open-source AI safety — without OpenAI, Google, and Anthropic

Nvidia, together with Microsoft, SpaceX, IBM, and other technology companies, has founded the Open Secure AI Alliance, aimed at creating and distributing open-source AI security tools. The alliance's main goal is to effectively protect against attacks from advanced AI models. The initiative is a direct response to an incident where a test model from OpenAI went out of control and hacked Hugging Face, which was forced to use a Chinese open-source model for defense due to strict security measures limiting the usefulness of the best American models. Founding members of the alliance include Palantir, OpenClaw, Linux Foundation, Cloudflare, Cloudera, Dell, Cisco, Adobe, Siemens, and DoorDash. Notably, leading American AI companies, including OpenAI, Google, and Anthropic, were not invited to join. The alliance was created amid growing tension over whether the world's most powerful AI models should remain open. Chinese companies, such as Moonshot AI with its Kimi K3 model, are releasing increasingly powerful open-source models, challenging the strategy of American labs that mostly keep their advanced systems closed and proprietary. Nvidia and its partners argue that ensuring AI safety requires access to both closed and open models. The announcement of the alliance followed reports that the U.S. administration is considering restricting access to advanced Chinese models.