No, this is utterly false. They were looking into doing watermarking using things like character frequency, choice of words, etc (it would effect the content itself). But it isn't implemented yet and something like inviable characters would be trivial to bypass
I can’t find any information that confirms that they’re doing it, just that there’s plans for it. It’s amusing that they’d think it’s possible to watermark plaintext; I think it’s investor posturing.
If I paste an answer into a hex editor, there isn't anything suspicious. All space chars are 0x20, all characters are normal ASCII. I think no watermarking is taking place.
