[antonkar]

Cool that they work in this direction. We need NVIDIA, OS providers to at least have robust updatable blacklists of bad AI models (probably agentic ones) else hackers and rogue states will make a botnet of viruses with agentic components.

We’ll have no time for “AI antiviruses”, agency explosion (grabbing GPUs, compute time, tools) precedes intelligence explosion.

OS providers and NVIDIA must have skin in the game, feel responsible for bad models that run on its GPUs/OSes, fund safety research, same way Apple feels responsible for apps in its App Store and viruses on MacOS