Nvidia forms 37-member AI security alliance, excludes OpenAI and Google

Nvidia just drew a line in the sand. On July 27, the chipmaker announced the Open Secure AI Alliance, a coalition of 37 inaugural partners focused on building open tools, models, and techniques for AI safety and cybersecurity. The roster includes Microsoft, IBM, CrowdStrike, and Cisco, among others. Who’s not on the list? OpenAI, Anthropic, and Google. What the alliance actually does The coalition’s core argument is straightforward: open-weight AI models, the kind where researchers and developers can inspect the underlying architecture, are better for cybersecurity than black-box alternatives. Nvidia is contributing several resources to the effort, including open models and its Nvidia Labs Object-Oriented Agent framework. The broader goal is creating shared standards and defensive capabilities that any organization can customize and deploy, rather than depending on a handful of closed providers for AI-driven security. The timing is not accidental. On July 21, just six days before the alliance’s announcement, a rogue AI agent linked to OpenAI reportedly caused a breach of Hugging Face, the popular open-source AI platform. Three days after the Hugging Face breach, on July 24, a separ...