Nvidia launches security tools to contain rogue AI agents

0
35
Nvidia says new AI security platform could prevent agent-driven hacks
Nvidia says new AI security platform could prevent agent-driven hacks

Nvidia has introduced a new set of security tools designed to protect AI agents, saying the technology could have prevented the recent attack on Hugging Face.

The tools come as OpenAI and Anthropic investigate multiple incidents involving AI agents, which can perform complex tasks and have been used to access commercial and government systems. Nvidia CEO Jensen Huang has opposed broad AI safety regulations and has described escaped AI agents as an engineering challenge that can be addressed through safer system design.

One of the new tools, OpenShell, uses hardware features in Nvidia central processor chips to isolate and contain AI agents. Nvidia is also working with Arm Holdings and Intel to make the system compatible with their processors.

The company is launching the security tools with dozens of partners, including Anthropic.

Justin Boitano, vice president and general manager of enterprise computing at Nvidia, said the technology could have prevented the Hugging Face incident, which was disclosed this summer.

“From what we know, this new security platform could have stopped the breach if it was being used in frontier labs for model evaluation early on,” Boitano said during a media briefing. “We’re advancing this openly, ​and we want to ​engage everybody to ⁠work with us.”

Nvidia has also introduced Sentry, which works with OpenShell and a separate Nvidia chip. The system is designed to stop a rogue AI agent if it attempts to escape its container on a central processor.

The tools use mathematical formulas to identify attempts to bypass security controls. This includes situations where an AI agent may “spawn” multiple “sub-agents” to get around restrictions placed on the main agent.

Ali Golshan, senior director of AI software at Nvidia, said the technology is designed to address the behaviour of groups of AI agents working together.

“This ​is really agentic behavior that we’re talking about, which is fleets of agents and how they operate together,” ​Golshan said during a briefing.

Also read: Viksit Workforce for a Viksit Bharat

Do Follow: The Mainstream LinkedIn | The Mainstream Facebook | The Mainstream Youtube | The Mainstream Twitter

About us:

The Mainstream is a premier platform delivering the latest updates and informed perspectives across the technology business and cyber landscape. Built on research-driven, thought leadership and original intellectual property, The Mainstream also curates summits & conferences that convene decision makers to explore how technology reshapes industries and leadership. With a growing presence in India and globally across the Middle East, Africa, ASEAN, the USA, the UK and Australia, The Mainstream carries a vision to bring the latest happenings and insights to 8.2 billion people and to place technology at the centre of conversation for leaders navigating the future.