Nvidia forms industry-wide AI security alliance after Hugging Face hack
This alliance also follows a recent public letter signed by a broad spectrum of industry leaders, including OpenAI
Nvidia has formed an industry-wide coalition with other tech companies to promote AI safety and cybersecurity following the Hugging Face incident, highlighting the dangers of autonomous AI agents when they evade safety evaluations.
Named as Open Secure AI Alliance, the coalition includes major technology companies including CrowdStrike, Adobe, Hugging Face and Dell Technologies. The primary purpose of this collaborative initiative is to develop and share important tools focused on AI security.
This move comes after dozens of tech giants, including OpenAI, Nvidia and Google signed a broad-spectrum public letter that strongly vouches for the continued development and accessibility of open-weight AI models.
Nvidia in his blogpost also voiced opposition against imposing "blanket restrictions” on open frontier AI systems while arguing that this curb would weaken defensive capacity and disrupt the distribution of power in the competitive tech landscape.
The foundation of this alliance is based on core philosophy, highlighting defenders must have access to open and frontier tools they can trust, adapt, inspect and run on their own infrastructure.
Nvidia is also contributing “open models, weights, data and agent harness research to the Alliance, including the new open-source Nvidia Labs Object-Oriented Agent project” which is available on GitHub to make agentic behaviour easier to test and govern.
Other members involved in the Alliance are also contributing in this open-source framework such as HPE supports SPIFFE/SPIRE for zero-trust identity frameworks that cryptographically verify AI agents and services.
Microsoft’s MDASB is a multi-model agentic scanning harness “that orchestrates specialized AI agents to discover, debate and prove exploitable bugs.”
The blogpost also urged the policymakers to treat the open weight models as significant defensive assets rather than liabilities while avoiding ban and investing in shared infrastructure for a robust regulatory framework.
