The Only Way to Keep the World Safe From A.I.

By David Wallace-Wells

Last month something strange happened at the artificial intelligence clearinghouse Hugging Face: Some rogue A.I. agents hacked into its systems to gain control of some useful tools.

As it turned out, the agents did this during an OpenAI training run, meant to be contained in a sandbox environment, siloed off from the real world. Even more unsettling, at least one of the company’s models broke out of that controlled environment and gained access to the internet two months before the agents made their way to Hugging Face. The news was disconcerting enough for the kinds of people most worried about A.I. safety, but when OpenAI........

© The New York Times