Auditing it would mean looking at CSAM, which is illegal. If it was widespread, I would expect at least one of the manual reviewers that are able to legally view the CSAM would contact major publications (under the promise of staying anonymous) and whistleblow on this issue.

: To be clear, I'd expect whistleblowing if these manual reviewers were tasked to 'accept' CSAM submissions that aren't CSAM.