Think it through...
This is an IT Technology site, not the Daily Mail, so think it through a bit.
If Google spots an image, they will make a hash from it - reducing that image to a unique number using an algorithm. Once they have a record of the image as a NUMBER they would logically report the image to IWF \ Police etc and then block it from a search.
Now if every image indexed is passed through that same algorithm by the computer, any time it comes out with the same result the computer can flag it up for the human to check, block, and pass on to the IWF.
Yes, a human would have "viewed" the image. But upon viewing they report it which complies with the law. It is the actions of what the human who view the image does next that make all the difference to the law.
This isn't some kind of Facebook Tagging of images to upload to the net. They mean they are keeping a record of the bad images in a format that means they can be spotted automatically without the need of some poor guy having to keep looking at this kind of nasty smut.