---
title: ">  AI companies still need to purchase at least one copy of anything..."
date: 2025-08-27
source: facebook
type: comment
context: "Ramez Naam's photo"
fb_link: "https://www.facebook.com/dyi/l/?l=AYOd7aI2NBk7TbnK9AqzGsb8N-pvDFqCwsWP3BPrE7T_MriZXezIFJHhLU55uSEwGkYP8EWYFgSXKq3M1EjbicuqoHIsSQrpfvIuz-85DiAuYJNknS7vKGKm-MEqFil_0GTGnE_Msnu1bZd_KhlB2mriWM8_4iW7QCDuON53XGdGB3Eoy4Kcsq-07VAF-aQ22vvt8qmjSmk&s=519"
---

# >  AI companies still need to purchase at least one copy of anything...

*August 27, 2025 — Comment Ramez Naam's photo*

>  AI companies still need to purchase at least one copy of anything copyright protected that they want to train on (which Anthropic didn't do). What?  That doesn't make sense.  What about libraries?