Nvidia vrea să împiedice agenții AI să „evadeze”. Cum îi va ține sub control

Nvidia has launched the Open Agent Safety Platform to help companies control AI agents' access to data and systems, addressing recent security incidents. This platform introduces external control mechanisms to prevent AI agents from bypassing restrictions. Key components include OpenShell, which secures environments and defines access limits, and Sentry, which monitors AI activity and can quickly isolate agents attempting to breach boundaries. Nvidia emphasizes that security must be enforced at the infrastructure level, not just through model instructions. The platform aims to enhance AI safety and will offer some software as open-source, allowing other developers to create security solutions. Major companies like Cisco and Microsoft are collaborating on this initiative.