You‘re not exposing business information, you‘re exposing client information. The information a system leaks might not be intrinsically valuable, but it can be used to deduce other data, especially over larger sets or time.

For example, by only scraping the date and author of an online newspaper‘s articles over a period of time, you can deduce when every author is typically on vacation. Compare that against every other author and you can find patterns indicating, say, workplace affairs.

Source: a talk by David Kreisel called SpiegelMining (in German), or at least what I remember.

Found an English translation of the mentioned video: https://www.youtube.com/watch?v=bYviBstTUwo