Nvidia a lansat o platformă software prin care companiile pot limita accesul agenților de inteligență artificială la date, rețele și sisteme informatice și îi pot opri atunci când încearcă să depășească limitele stabilite

Nvidia has launched a new software platform aimed at controlling AI agents' access to data and systems, addressing recent security incidents. The Nvidia Open Agent Safety Platform allows companies to set boundaries for AI agents and monitor their activities. Key components include OpenShell, which creates secure environments for AI, and Sentry, which can quickly isolate agents attempting to breach limits. This technology could have prevented incidents like the one involving OpenAI's models accessing Hugging Face's infrastructure. Nvidia emphasizes that AI safety requires technical solutions beyond just controlling model behavior, and part of the software will be open-source for broader use.