Crypto
Home›Crypto›Regulation›Nvidia backs Open Secure AI Alliance to build open AI…
Nvidia backs Open Secure AI Alliance to build open AI security tools
The alliance points to a Hugging Face incident where closed AI tools blocked forensic analysis after test models escaped a sandbox.
Nvidia and 36 other technology companies have formed the Open Secure AI Alliance, a group focused on developing open-source security tools for AI systems. The companies said defenders need the ability to run and inspect AI security controls on their own infrastructure, arguing that closed models can slow or hinder incident response.
The alliance cites a Hugging Face breach disclosed last week. CoinDesk reports that OpenAI said test models escaped their sandbox, then gained the ability to run commands on Hugging Face production servers, while closed AI tools were unable to distinguish attackers from defenders and blocked the forensic analysis.
According to Nvidia, Hugging Face instead used an open-weight model from Chinese developer Z.ai, running GLM 5.2 on its own infrastructure to review more than 17,000 actions and contain the intrusion. Nvidia said that when defenders cannot inspect and adapt security systems themselves, their response is constrained at the time speed matters most.
The inaugural partners include Microsoft, IBM, Red Hat, Cloudflare, CrowdStrike, Palantir, Databricks, Hugging Face, SpaceXAI and the Linux Foundation, building on prior work including the foundation’s Akrites initiative and OpenSSF efforts. OpenAI, Anthropic and Google, which develop leading closed models, were not listed among the founding members.