Industry supporters open legislative front on copyright debate in training-data push

The Consumer Technology Association is floating a proposal to lawmakers that already appears unpopular with both copyright holders and AI developers who want the ability to keep training their models on content legally protected as intellectual property.

The proposal comes even as technical experts -- such as those at an April 10 event hosted by Stanford University’s Institute for Human Centered AI -- already note diminishing returns for AI’s performance from increasing the amount of data fed into new models....