The author's bias - it's different for each specific author. We should not pretend that there are moderators without bias, each AI-driven moderation tool inherits the bias of its human author.
The LLMs that power all that are "aligned", that is, they're subjected to manipulation to install specific bias in them, and so on.
