I wouldn't dare call it mislabelling since there is no precedent yet. Just the other day a judge ruled AI generated CSAM is still CSAM. If it can be proven beyond a doubt that an AI output comes from copyrighted works without proper license, will that AI violate the copyright? Also, will AI count as derivatives work from the training material or will it be treated like software compiler? I think a lot of our current legal framework is not up to speed to answer those questions. So I would not call it useless nor misleading.
Also, lemmy doesn't have EULA as far as I am aware of so the license of the content hosted on the instance is by default unlicensed. The user just notifies that to whoever wants to use their comment for whatever purpose, must abide by those licenses.