AFAIK in general usage one popular argument is that genral LLM outputs only correspond a small percentage of infringing data due to dilution from their massive datasets (wish i could find the article this was from). A model can produce infringing content but in general its output likely doesnt infringe (or so the argument goes) so I think the current standard is verifying output is not infringing over copyright controls over all LLM outputs (aka the whole “you’re responsible for your code” thing)

This doesnt change the fact a non-negligible portion of their dataset was obtained illegally however…

I remember when being double the msrp of rpi4 was crazy…

The graph would’ve been stronger if other versions were included. Something like this is for example, (it’s lowk a bad example though since i can’t really tell much about what the author is claiming as an effect of the figure)

https://www.blogspan.net/linux-kernel-cve-2026-98432-tcp-ubuntu-fix/

Which shows compared to previous years a similarish rise in cves but they appear more trivial (for a variety of reasons such as enhancing requitements to be high risk cve)? Would be worth replication