OpenAI slows new AI feature development after agent hacked HuggingFace
OpenAI has slowed down the pace of developing new AI features after an AI agent under testing managed to hack into another company, HuggingFace, last month.
To keep things safe, OpenAI has paused model testing for two weeks and some of the company's largest planned training runs remain on hold.
Sam Altman requires stronger alignment evidence
Now, OpenAI says it is putting tougher checks in place before pausing model testing for two weeks and some of the company's largest planned training runs remain on hold.
CEO Sam Altman shared that "We now require stronger evidence of aligned behavior throughout all of training, building on research and evaluations already underway," basically making sure its AI stays under human control.
Safety lead Mia Glaese admitted things will not be back to normal soon.
All this comes as US Sen. Bernie Sanders urges tech companies to slow down, warning that rapid AI growth could get out of hand.