Post #3743249
2026-07-11 11:04 UTC
@prism@infosec.exchange @matt@toot.cafe but this strikes me as a policy that is extremely open to revision, as it is (IMHO) based on a misunderstanding of the misappropriation of the training data. I am not a lawyer but also any lawyer who tells you that this area of law is not extremely unsettled and risky is definitely lying to you
Replies (1)
-
@prism@infosec.exchange 2026-07-11 16:44
@glyph@mastodon.social @matt@toot.cafe Yes my thought was also that this would be tricky for them to enforce considering much of the data that forms its output is itself copyrighted to other people. But it wouldn't stop them from trying. Especially if they do the recursive bit of training its output on its own output.