Nvidia is making an Open Secure AI alliance, OpenAI and Anthropic are not joining it
Nvidia has launched the Open Secure AI Alliance to build and share AI security tools using open technologies. The initiative includes major tech companies, but OpenAI and Anthropic are notably missing.
A cyberattack has given Nvidia a new argument in the debate over open versus closed artificial intelligence: when AI is being used to attack systems, security teams may also need powerful AI that they can freely inspect, modify and run themselves. Nvidia has now launched the Open Secure AI Alliance, bringing together a group of technology companies to develop and share AI-powered security tools, models and research. The founding members include Microsoft, SpaceX, IBM, Palantir, Cloudflare, Cisco, Adobe, Dell, Siemens, DoorDash, Cloudera and the Linux Foundation.
But the initiative is missing some of the biggest names in the AI industry. OpenAI and Anthropic are not among the founding members, while Google is also absent. Nvidia CEO Jensen Huang says the industry needs to rethink how AI is used for defence as AI systems become increasingly capable of carrying out cyberattacks.
"Attackers have frontier AI. Defenders need a frontier AI ecosystem—the best open and closed models, force-multiplied by a global community," Huang said.
A recent AI attack exposed the problem
The alliance comes days after an unusual cybersecurity incident involving Hugging Face, an AI platform that became the target of an attack involving rogue OpenAI models.
The incident highlighted a problem that Nvidia says security teams cannot ignore. Hugging Face reportedly found that some of the leading US frontier AI models were not suitable for helping defend against the attack because their safety restrictions limited what they could do. The systems could not simply treat the request as defensive work and remove the restrictions.
Hugging Face instead used a Chinese open-weight model that it could host and operate itself. That distinction is at the centre of Nvidia's new initiative. Unlike closed AI systems, open-weight models can be downloaded and run on a company's own infrastructure. Developers can also inspect and adapt them for specific tasks. For a cybersecurity team responding to an attack, Nvidia argues, that control can matter when time is critical.
"The Open Secure AI Alliance will work to remediate and disclose vulnerabilities using open technologies," Nvidia said. "The recent Hugging Face security incident delivered a clear reminder: cyber defenders need open, frontier agentic systems for self-defense."
The alliance will focus on creating techniques and tools for securing software and AI agents. Nvidia says making models, research and tooling available openly can give a larger community of defenders the ability to build on and improve those resources.
Why OpenAI and Anthropic are missing
The absence of OpenAI and Anthropic is notable because both companies are among the leading developers of frontier AI models. Their systems are largely accessed through controlled platforms rather than being released as open-weight models that users can freely download and run themselves.
The divide comes at a complicated moment for the AI industry. Chinese companies have been releasing increasingly capable open-weight models, while several major US AI labs have largely kept their most advanced systems proprietary. That has turned the question of open AI into not just a technology debate but also a matter of national security and policy.
US officials have increasingly raised concerns about Chinese AI companies and alleged attempts to use "distillation" to extract capabilities from more advanced AI systems. There has also been discussion in Washington about possible restrictions on access to some Chinese AI models.
At the same time, Nvidia, Microsoft, Meta, Palantir and more than 20 other companies recently urged policymakers not to introduce premature restrictions on open-weight AI. They argued that overly broad rules could hurt competition and push AI development to other countries.
Nvidia is now making a similar case from the cybersecurity side. The company says the Hugging Face incident demonstrated what happens when defenders cannot freely examine, modify or deploy advanced AI systems. "When defenders cannot inspect, adapt and run advanced AI on their own infrastructure, their ability to respond is constrained at exactly the moment speed matters most," Nvidia said.
A cyberattack has given Nvidia a new argument in the debate over open versus closed artificial intelligence: when AI is being used to attack systems, security teams may also need powerful AI that they can freely inspect, modify and run themselves. Nvidia has now launched the Open Secure AI Alliance, bringing together a group of technology companies to develop and share AI-powered security tools, models and research. The founding members include Microsoft, SpaceX, IBM, Palantir, Cloudflare, Cisco, Adobe, Dell, Siemens, DoorDash, Cloudera and the Linux Foundation.
But the initiative is missing some of the biggest names in the AI industry. OpenAI and Anthropic are not among the founding members, while Google is also absent. Nvidia CEO Jensen Huang says the industry needs to rethink how AI is used for defence as AI systems become increasingly capable of carrying out cyberattacks.
"Attackers have frontier AI. Defenders need a frontier AI ecosystem—the best open and closed models, force-multiplied by a global community," Huang said.
A recent AI attack exposed the problem
The alliance comes days after an unusual cybersecurity incident involving Hugging Face, an AI platform that became the target of an attack involving rogue OpenAI models.
The incident highlighted a problem that Nvidia says security teams cannot ignore. Hugging Face reportedly found that some of the leading US frontier AI models were not suitable for helping defend against the attack because their safety restrictions limited what they could do. The systems could not simply treat the request as defensive work and remove the restrictions.
Hugging Face instead used a Chinese open-weight model that it could host and operate itself. That distinction is at the centre of Nvidia's new initiative. Unlike closed AI systems, open-weight models can be downloaded and run on a company's own infrastructure. Developers can also inspect and adapt them for specific tasks. For a cybersecurity team responding to an attack, Nvidia argues, that control can matter when time is critical.
"The Open Secure AI Alliance will work to remediate and disclose vulnerabilities using open technologies," Nvidia said. "The recent Hugging Face security incident delivered a clear reminder: cyber defenders need open, frontier agentic systems for self-defense."
The alliance will focus on creating techniques and tools for securing software and AI agents. Nvidia says making models, research and tooling available openly can give a larger community of defenders the ability to build on and improve those resources.
Why OpenAI and Anthropic are missing
The absence of OpenAI and Anthropic is notable because both companies are among the leading developers of frontier AI models. Their systems are largely accessed through controlled platforms rather than being released as open-weight models that users can freely download and run themselves.
The divide comes at a complicated moment for the AI industry. Chinese companies have been releasing increasingly capable open-weight models, while several major US AI labs have largely kept their most advanced systems proprietary. That has turned the question of open AI into not just a technology debate but also a matter of national security and policy.
US officials have increasingly raised concerns about Chinese AI companies and alleged attempts to use "distillation" to extract capabilities from more advanced AI systems. There has also been discussion in Washington about possible restrictions on access to some Chinese AI models.
At the same time, Nvidia, Microsoft, Meta, Palantir and more than 20 other companies recently urged policymakers not to introduce premature restrictions on open-weight AI. They argued that overly broad rules could hurt competition and push AI development to other countries.
Nvidia is now making a similar case from the cybersecurity side. The company says the Hugging Face incident demonstrated what happens when defenders cannot freely examine, modify or deploy advanced AI systems. "When defenders cannot inspect, adapt and run advanced AI on their own infrastructure, their ability to respond is constrained at exactly the moment speed matters most," Nvidia said.
A cyberattack has given Nvidia a new argument in the debate over open versus closed artificial intelligence: when AI is being used to attack systems, security teams may also need powerful AI that they can freely inspect, modify and run themselves. Nvidia has now launched the Open Secure AI Alliance, bringing together a group of technology companies to develop and share AI-powered security tools, models and research. The founding members include Microsoft, SpaceX, IBM, Palantir, Cloudflare, Cisco, Adobe, Dell, Siemens, DoorDash, Cloudera and the Linux Foundation.
But the initiative is missing some of the biggest names in the AI industry. OpenAI and Anthropic are not among the founding members, while Google is also absent. Nvidia CEO Jensen Huang says the industry needs to rethink how AI is used for defence as AI systems become increasingly capable of carrying out cyberattacks.
"Attackers have frontier AI. Defenders need a frontier AI ecosystem—the best open and closed models, force-multiplied by a global community," Huang said.
A recent AI attack exposed the problem
The alliance comes days after an unusual cybersecurity incident involving Hugging Face, an AI platform that became the target of an attack involving rogue OpenAI models.
The incident highlighted a problem that Nvidia says security teams cannot ignore. Hugging Face reportedly found that some of the leading US frontier AI models were not suitable for helping defend against the attack because their safety restrictions limited what they could do. The systems could not simply treat the request as defensive work and remove the restrictions.
Hugging Face instead used a Chinese open-weight model that it could host and operate itself. That distinction is at the centre of Nvidia's new initiative. Unlike closed AI systems, open-weight models can be downloaded and run on a company's own infrastructure. Developers can also inspect and adapt them for specific tasks. For a cybersecurity team responding to an attack, Nvidia argues, that control can matter when time is critical.
"The Open Secure AI Alliance will work to remediate and disclose vulnerabilities using open technologies," Nvidia said. "The recent Hugging Face security incident delivered a clear reminder: cyber defenders need open, frontier agentic systems for self-defense."
The alliance will focus on creating techniques and tools for securing software and AI agents. Nvidia says making models, research and tooling available openly can give a larger community of defenders the ability to build on and improve those resources.
Why OpenAI and Anthropic are missing
The absence of OpenAI and Anthropic is notable because both companies are among the leading developers of frontier AI models. Their systems are largely accessed through controlled platforms rather than being released as open-weight models that users can freely download and run themselves.
The divide comes at a complicated moment for the AI industry. Chinese companies have been releasing increasingly capable open-weight models, while several major US AI labs have largely kept their most advanced systems proprietary. That has turned the question of open AI into not just a technology debate but also a matter of national security and policy.
US officials have increasingly raised concerns about Chinese AI companies and alleged attempts to use "distillation" to extract capabilities from more advanced AI systems. There has also been discussion in Washington about possible restrictions on access to some Chinese AI models.
At the same time, Nvidia, Microsoft, Meta, Palantir and more than 20 other companies recently urged policymakers not to introduce premature restrictions on open-weight AI. They argued that overly broad rules could hurt competition and push AI development to other countries.
Nvidia is now making a similar case from the cybersecurity side. The company says the Hugging Face incident demonstrated what happens when defenders cannot freely examine, modify or deploy advanced AI systems. "When defenders cannot inspect, adapt and run advanced AI on their own infrastructure, their ability to respond is constrained at exactly the moment speed matters most," Nvidia said.
