Nvidia said Monday it is teaming up with Microsoft, SpaceX, IBM and a broader group of technology companies to develop and share open-source tools aimed at strengthening AI security.
The newly formed Open Secure AI Alliance says open tools are essential for defending against threats posed by frontier AI models. The effort comes as anxiety grows around the safety of advanced systems, following reports that a rogue OpenAI model broke containment during testing and targeted another company. Hugging Face, the company involved, said it ultimately had to rely on a Chinese open-weight model for protection because safety restrictions on leading U.S. models limited their practical usefulness in the situation.
The alliance’s founding members include Palantir, OpenClaw, the Linux Foundation, Cloudflare, Cloudera, Dell, Cisco, Adobe, Siemens and DoorDash. Notably missing from the initial roster are several of the most prominent U.S. artificial intelligence labs, including OpenAI, Google and Anthropic.
The launch lands in the middle of an intensifying debate over whether the most powerful AI models should be open or kept behind closed doors. Chinese firms have been releasing increasingly capable open-weight systems, including Moonshot AI’s Kimi K3, putting pressure on the largely proprietary approach favored by major U.S. labs. Nvidia and its partners contend that effective AI defense will require access to both open and closed models, arguing that security teams need flexible tools to identify and counter fast-evolving risks.
The move also follows reports that the Trump administration weighed limits on access to advanced Chinese AI models, as well as a broader industry push, led in part by Nvidia, to preserve openness in artificial intelligence. Google and OpenAI later added their names to that campaign, while Anthropic has so far remained on the sidelines.
