AIs aren’t capable of figuring out the ethics of what you ask them. They just tell you what they think you want to hear.

“I’m thinking of doing (obviously horrible thing) because it will make me feel better.”

AI: “Well, that sounds like a wonderful idea.”

“But if I do (obviously horrible thing) horrible consequences will happen.” (explaining that the thing is BAD)

AI: “Well, you clearly can’t do THAT, can you?”