> Why is the automatic assumption that it was malicious?

Because the business model is malicious.

Meta has, quite intentionally, wedged itself in-between large amounts of human interaction. They know some people are malicious. They know the machine will frequently make the wrong call, because the auto-moderator can neither reason, nor account for context.

But hiring enough moderators to run moderation through the computer, provide a default "suggested" action, and having a human approve or deny, would be too expensive. It would demolish their profit, and may even pull Meta from black to red, financially. Their business model relies on making the wrong call, and having outrage bring it to the attention of actual Meta staff to correct it.

Meta does not get to blame the computer they put in charge of moderation instead of hiring adequate staff; computers can never be held accountable, so accountability moves up the chain to the people that programmed the computer and left it in charge.