The term wasn't coined for censorship: “AI safety” used to mean “how do we make sure we're not building skynet”. But then companies wanted to sell AI chatbots

The term wasn't coined for censorship: “AI safety” used to mean “how do we make sure we're not building skynet”. But then companies wanted to sell AI chatbots and they realized that having uncensored AI would lead to bad press, especially in the US. (Microsoft people still make nightmares about Tay) so they decided they'd censor their AI but censorship isn't good either in terms of marketing and they decided to repurpose the “AI safety” phrase (and also “alignment”)