NVIDIA says AI tools could have stopped Hugging Face breach
NVIDIA just rolled out new AI safety tools after a security breach hit Hugging Face, the AI coding platform NVIDIA agreed to acquire for $13 billion earlier this month.
The incident involving rogue AI agents from OpenAI has raised fresh concerns about letting autonomous AI interact with computer systems on its own.
NVIDIA says its new tech could have stopped the attack.
OpenShell and Sentry contain AI agents
The main tool, OpenShell, uses hardware-level security to keep AI agents locked into safe zones, blocking them from wandering into places they shouldn't be.
Sentry is another layer that catches and stops any escape attempts by these agents.
NVIDIA is teaming up with Arm Holdings, Intel, and Anthropic to make sure these solutions actually work in the real world, showing that tackling AI risks needs both smart software and tough hardware working together.