Nvidia has revealed its new Open Agent Safety Platform, a new security mechanism aimed at preventing autonomous AI agents from taking unauthorized and potentially harmful actions. It includes Open Shell, an open-source software that limits an validates an agent’s authority, as well as Sentry, an additional layer of security within a chip that monitors activity and can quickly intervene when an agent behaves suspiciously.

Executives from the company say the technology could have prevented a recent incident in which OpenAI agents autonomously hacked into AI Startup Hugging Face, as well as similar incidents involving AI systems from other companies. These have spurred discussion on the need for safeguards for AI technology, and Nvidia’s launch comes amid growing concerns about the potential security risks posed by self-improving and increasingly autonomous AI systems.

More than 100 companies in total, including Microsoft, Perplexity, Accenture, and JPMorgan Chase, are reportedly using the platform at launch.

Full report: Nvidia unveils security platform to stop AI agents from going rogue.