You can read about it here, it is relatively short, coherent, and worth reading over: 

https://ico.org.uk/about-the-ico/what-we-do/our-work-on-artificial-intelligence/generative-ai-second-call-for-evidence/

It sets out some positions for consideration around separating purposes of training from purposes of application building. I think it's a good start in the area of looking at transparency around model building.