Nvidia and Microsoft, together with SpaceX, IBM, and a dozen other technology firms, announced a joint initiative to develop open-source tools for AI security. The coalition aims to create shared defenses against potential threats from advanced AI systems, responding to growing concerns about safety in the rapidly evolving field. Founding members also include Palantir, Cloudflare, Dell, Cisco, Adobe, Siemens, and DoorDash. Conspicuously missing from the list are the leading US AI labs: OpenAI, Google, and Anthropic.

The initiative arrives amid an intensifying debate over whether the most powerful AI models should remain proprietary or be released openly. Chinese companies have been distributing increasingly capable open-weight models, challenging the closed approach favored by many American labs. Proponents of openness argue that security requires broad access to model internals, enabling defenders to anticipate and counter attacks. The new alliance stresses that both open and closed models must be part of a comprehensive security strategy.

The announcement follows reports that the Trump administration considered restricting access to Chinese AI models, and an industry-led campaign—again spearheaded by Nvidia—urging the government to maintain openness. Google and OpenAI eventually signed that letter, though Anthropic has not. The formation of this security group signals that major technology players are moving to coordinate defenses in an environment where AI threats are no longer hypothetical.

Source: The Verge