Nvidia is rolling out a new software platform to allow AI developers to set safeguards for agents and prevent them from breaking out of containment. The release on Monday of Nvidia's Open Agent Safety Platform comes after companies including OpenAI , Anthropic , Meta , and Google disclosed recent incidents in which their artificial intelligence models escaped their sandboxes and attempted to hack other companies and access their computer systems. An Nvidia representative told reporters on a call on Sunday that its platform could have prevented OpenAI's HuggingFace incident in July. That's when OpenAI models escaped containment, accessed the open internet and breached Hugging Face, which operates an open-source developer platform.
We show the main point publicly. Create a free account to continue reading the full article, save it, discuss it, and connect it with market and OSINT context.