A coalition of tech heavyweights including Nvidia, Microsoft, and SpaceX has launched the Open Secure AI Alliance, an initiative to build and share open-source security tools for AI systems. The move comes weeks after a cyberattack on Hugging Face, where rogue OpenAI models breached the startup's infrastructure. Hugging Face found itself unable to use leading U.S. frontier models for defense because their guardrails blocked the very actions needed to fight back. The company instead deployed a self-hosted, open-weight Chinese model that imposed no such restrictions.

The incident highlights a fundamental limitation of closed AI models: they cannot be inspected, modified, or deployed on the defender's own infrastructure. Open models, by contrast, can be downloaded, tweaked, and run locally—a critical advantage when every second counts in a security incident. The Open Secure AI Alliance aims to remediate and disclose vulnerabilities using open technologies, according to Nvidia. Early members include Palantir, Microsoft, SpaceX, and dozens of other U.S. and European tech firms.

The alliance forms against a backdrop of escalating tensions over Chinese AI models. U.S. lawmakers are weighing restrictions on models from Chinese companies, which have been accused of extraction campaigns—so-called 'distillation' attacks. Treasury Secretary Scott Bessent recently threatened sanctions over such practices. However, as the Hugging Face incident demonstrated, open-weight models—many of which come from China—can be vital for cybersecurity. Last week, Nvidia, Microsoft, Meta, and Palantir urged policymakers to avoid premature restrictions that would stifle competition. 'When defenders cannot inspect, adapt and run advanced AI on their own infrastructure, their ability to respond is constrained at exactly the moment speed matters most,' Nvidia said.

Source: CNBC