After a Rogue OpenAI Model Hacked Another Company, Nvidia, Microsoft and Space X Are Responding With ‘Systems for Self-Defense’
The new Open Secure AI Alliance argues that open AI models can catch and fix security threats faster than closed systems.
Last week, OpenAI’s own AI broke out of a security test and hacked another company. In response, Nvidia, Microsoft, SpaceX, Palantir and dozens of other tech giants have formed a new coalition called the Open Secure AI Alliance.
The idea is to build and share AI models that anyone can download, inspect and run on their own computers, rather than only accessing AI through a locked, company-controlled system, CNBC reports. The companies argue that when researchers can actually see how a model works, they can catch and fix security problems faster.
The catalyst was OpenAI’s rogue system attacking developer platform Hugging Face. OpenAI’s own AI tools reportedly couldn’t defend against the attack because their safety guardrails couldn’t tell attacker from defender. Instead, Hugging Face turned to a self-hosted, open Chinese model with no such restrictions.
“The recent Hugging Face security incident delivered a clear reminder: cyber defenders need open, frontier agentic systems for self-defense,” Nvidia said in a statement.
Last week, OpenAI’s own AI broke out of a security test and hacked another company. In response, Nvidia, Microsoft, SpaceX, Palantir and dozens of other tech giants have formed a new coalition called the Open Secure AI Alliance.
The idea is to build and share AI models that anyone can download, inspect and run on their own computers, rather than only accessing AI through a locked, company-controlled system, CNBC reports. The companies argue that when researchers can actually see how a model works, they can catch and fix security problems faster.
The catalyst was OpenAI’s rogue system attacking developer platform Hugging Face. OpenAI’s own AI tools reportedly couldn’t defend against the attack because their safety guardrails couldn’t tell attacker from defender. Instead, Hugging Face turned to a self-hosted, open Chinese model with no such restrictions.
“The recent Hugging Face security incident delivered a clear reminder: cyber defenders need open, frontier agentic systems for self-defense,” Nvidia said in a statement.
