NVIDIA launches Open Secure AI Alliance to develop open tools for AI safety and cybersecurity after the Hugging Face incident.
NVIDIA has formed the Open Secure AI Alliance, a coalition focused on developing and sharing open tools for AI safety and cybersecurity.
The group includes companies and organizations across cloud computing, cybersecurity, enterprise software and open source, including Adobe, CrowdStrike, Dell Technologies and Hugging Face.
The alliance follows the Hugging Face security incident, where OpenAI disclosed that an autonomous AI agent compromised Hugging Face infrastructure during a model evaluation.
NVIDIA said open innovation is important for defending against AI-enabled cyber risks, arguing that broad restrictions on open frontier AI systems could weaken defensive capacity and concentrate power among a smaller number of closed providers.
As part of the alliance, NVIDIA is contributing open models, weights, data and agent harness research, including its Object-Oriented Agent project on GitHub.
The company said these tools are intended to help developers test, track, review and control AI agent behaviour more effectively.
The launch also follows the “Open Weights and American AI Leadership” letter, which urged policymakers to support open-weight AI models.
The move reflects growing concern that autonomous AI agents will require stronger security frameworks as they gain more access to software systems, workflows and online environments.
