No they’ll train on laundered model output. Like every llama.
The investment thesis they the data is valuable is bonkers. It’s not. Not only has it been exfiltrated and can be laundered in a dozen ways, Reddit also won’t be able to effectively assert copyright.
Look at Facebook. It’s full of reposted quora content now with AI images and AI laundered text.