I find it weird how I don’t see any mention of the TDM exceptions (“Text and Data Mining”) that already explicitly allows AI companies to train on copyrighted data, in some cases even allowing them to ignore opt-outs (such as in research institutions). This is already implemented in the UK, EU, Japan and Singapore.
It seems to me that the online discourse is very US-centric, thinking that the AI regulatory battles are in the future, when in some other countries it’s already over.