Copyright law should allow building statistical models over copyrighted work (without reproducing it) so AI systems can become smarter; China will gain AI advantage if the US restricts model training on complete data.

normativepending

Speaker

Ben Horowitz

Evidence Quote

Can you build a statistical model over copyrighted work without reproducing it? We think that's important because China won't respect copyrights and we'll have weaker AI if we can't train on complete data.

Source

Ben Horowitz and David Solomon: The Sweetest Macro Spot in 40 Yearsa16z
Created: 8/12/2026, 10:05:59 PM

My Notes

Loading notes...