Reposted by T.J. Thomson
RMIT's @tjthomson.bsky.social tells me he doesn't believe AI text watermarks will change ordinary people's use of LLMs for basic tasks "at this early stage".
But those with expertise may try to "cover their tracks and game the system" by editing text to weaken the mark, or they'll use other models.
ia.acs.org.au
Anthropic's Claude to watermark AI-generated text
OpenAI also agrees to transparency rules for ChatGPT.