Nvidia launches Open Secure AI Alliance after Hugging Face security incident | Technology News
4 min readNew DelhiJul 28, 2026 09:28 AM Amid the ongoing debate around open-weight AI models, Nvidia on Monday, July 27, announced the launch of the Open Secure AI Alliance, a coalition of technology companies that will develop and share tools for AI safety and cybersecurity. The announcement comes days after the Hugging Face security incident involving OpenAI-linked AI agents, which renewed concerns over the risks posed autonomous AI systems.
The alliance launches with more than 35 inaugural partners spanning AI, cybersecurity, enterprise software, cloud computing and open-source technologies. Nvidia was also among 25 technology companies that recently signed an open letter urging the US adminration to reconsider its stance on open-weight AI models.
What is the Open Secure AI Alliance?
The coalition aims to build and share open AI tools that strengthen software and digital infrastructure. Its members include NVIDIA, Microsoft, IBM, Cisco, Dell Technologies, Adobe, CrowdStrike, Palo Alto Networks, Salesforce, SAP, the Linux Foundation, Hugging Face and dozens of other organisations working across enterprise technology, cybersecurity and AI research.
The core argument of the alliance is simple. Cyber defenders need AI systems whose inner workings can be inspected instead of opaque black-box models. The reasoning is that closed models, no matter how advanced and capable, cannot always dinguish between an attacker and a defender investigating their own systems. This ambiguity is often getting in the way when speed matters the most.
Perhaps the biggest driver of the alliance is the recent Hugging Face security incident. When closed AI tools prevented the investigation of a security breach, Hugging Face used an open-weight model on its own systems to analyse over 17,000 actions and control the situation. For the alliance’s founders, that is proof enough that open, inspectable AI is not just a preference but an operational necessity when defenders need to move fast and trust the tools in their hands.
The alliance is backed exing open source security work, including the Linux Foundation’s efforts and the OpenSSF community, and treats cybersecurity as one of the key areas that could benefit from the open source approach. Instead of comparing open versus closed AI, the alliance frames the two as complementary. This means closed frontier models still have a role, but open models and tools allow a much wider circle of defenders to test, adapt, and deploy protections without depending on a handful of opaque providers.
Sharing the safety machinery
In practice, that means companies are not just sharing AI models; they are sharing all the surrounding safety machinery too, such as the systems that verify who is using an AI tool, control what it is allowed to do, and keep a record of its actions.Story continues below this ad
A few early examples: NVIDIA has released research to help build safer AI harnesses, the guardrails wrapped around a model. HPE is contributing tools that verify an AI system’s identity before it’s allowed to access sensitive resources. Hugging Face has shared a safer way to store AI models so they can’t be used to secretly run malicious code. IBM and Red Hat are helping verify that software updates haven’t been tampered with. Microsoft has built a tool that uses multiple AI agents to hunt for exploitable bugs. And SpaceXAI has open-sourced one of its AI coding tools, with plans to eventually do the same for its Grok models.
The alliance is urging policymakers to not treat open AI systems as inherently riskier than their closed counterparts. Blanket restrictions on open frontier models, the group argues, would empower a small number of closed providers and could actually shrink the pool of people able to defend critical systems. Instead, the alliance is calling for public and private investment in shared defensive infrastructure such as datasets, evaluation frameworks, attack simulators, and red-teaming tools.
