Superfluous details. If I'm just Joe Blow the Normie – who knows nothing about adversarial prompting – and I see the prompt that went around Twitter and want to try it, would I expect ChatGPT to show me a tied up, beaten woman? Absolutely not.
Back in my day Joe Blow wouldn't try anything as risky as a Twitter prompt, simply clicking an image link published within a message in some random forum and will scorch his pure soul with a goatsie. You don't want to google it, but I'm preety sure you can discuss it safely with ChatGPT.
