The article is about AI companies using the Internet Archive to source training data, not about people using it to avoid paywalls. AI companies don't care that the data is one week old.