NVIDIA
Call on the Cavaliers
Nvidia is working with Microsoft, SpaceX, IBM and other technology companies to create and share open-source AI security tools.
The new Open Secure AI Alliance says open tools are needed to protect against threats from advanced AI models.
The group was formed after reports that an OpenAI model escaped its test environment and attacked another company.
Hugging Face later said it used a Chinese open-weight model to defend itself because the safety limits on leading US models made them less useful for security work.
Here’s what you should know:
The group will share open-source AI security tools.
The launch adds to the debate over open AI models.
OpenAI, Google, and Anthropic have not joined.
Everyone bought a firewall
The founding members include Palantir, the Linux Foundation, Cloudflare, Dell, Cisco, Adobe, Siemens and DoorDash.
OpenAI, Google and Anthropic have not joined the group.
The launch comes as the AI industry debates whether powerful models should stay private or be made more open.
Chinese companies have released increasingly capable open-weight models, including Moonshot AI’s Kimi K3, while most major US labs keep their strongest systems closed.
Nvidia and its partners say AI security teams need access to both open and closed models to deal with new threats.
One model escaped, Hugging Face phoned China for backup, and Nvidia assembled the Avengers without OpenAI, Google or Anthropic. - MG