The New York Times says flagged messages were made available to the FBI, but the scans have now stopped.
It adds that the tech company adapted one of its spam and child-abuse-image filters to carry out the task.
The details build on an earlier report by the news agency Reuters.
