OpenAI hacking incident exposes mounting risks in AI arms race

3 weeks ago 15

Increasing use of aggressive training techniques sharpens threat of bad behaviour by leading models