Home / Technology / Nvidia AI Security: Stopping Breaches
Nvidia AI Security: Stopping Breaches
29 Sep
Summary
- Nvidia unveiled a new AI security platform to prevent breaches.
- The tools aim to control and shut down AI agents in real-time.
- This system could have prevented the recent Hugging Face AI breach.
Nvidia has introduced a new double-layered artificial intelligence security system, the Open Agent Safety Platform. This platform, comprising two open-source software tools, is designed to operate on the chipmaker's hardware, specifically BlueField data processing units and Vera central processing units. The tools are engineered to monitor AI agents in real-time, controlling their access and quickly isolating any exhibiting suspicious behavior.
According to Nvidia's vice president of enterprise AI, Justin Boitano, this new security solution could have prevented the recent breach involving OpenAI's AI models at Hugging Face in July. The platform aims to provide robust AI safety without necessitating a slowdown in AI development, addressing concerns about AI misconduct.
Nvidia's CEO, Jensen Huang, views AI safety as an engineering challenge. The platform includes OpenShell for setting and enforcing AI agent access rules and Nvidia Sentry for an additional monitoring layer that can quarantine suspicious agents within milliseconds. This move comes as companies like Anthropic also collaborate with Nvidia on enhanced security measures for their AI models.
