Hm? it doesn't change its behavior based on interactions with users. It only changes when openai changes it.

Look at all the safety controls they had to add very early on, and how easily people bypassed them for awhile.

It's a mistake to call them safety controls, they're public relations controls. Nothing they did made it more safe for anyone, it made it less likely to embarrass them in the press.

AI companies that can't differentiate polishing their public image from safety should probably rake pretty highly on our risk of AI risk sources. :)