[aeternum]

It's also a very questionable way to source training data.

Most bots should be programmed to avoid violating Twitter rules to ensure they are not banned.

Maybe Bot Sentinel's model is simply finding users that prefer not to follow authority/rules.