See https://technomancers.ai/japan-goes-all-in-copyright-doesnt-apply-to-ai-training/ (which cites a Japanese-langauge source https://go2senkyo.com/seijika/122181/posts/685617 that I can't read).
According to this, Japan won't enforce copyright protections on data used to train AI models.
Will this change within the decade?
Resolves YES if either:
(1) Both of these:
(a) A rights holder uses Japanese copyright law to prevent an AI model from being trained, used, or distributed
(b) The legal claim is based on the argument that model was trained on copyrighted data, which the model creator did not have permission to use
(2) Japan passes a law / regulation / judicial ruling that plainly enables the sort of claim described in (1)
NO if there's no such case by 2030.
Ability to enforce payment of royalties counts under 1a. An AI model creator who voluntarily pays royalties does not count.
