Google sheds light on its child porn detector after sex offender arrest

22 Aug 2014 • engadget.com

While the spokesperson didn’t talk about it in detail, that technology might be what the company supposedly developed last year: a system that can trawl the web and email accounts for child sex images. Gmail’s rep made clear, however, that Mountain View only uses the system to detect child porn – it can’t actually parse emails that contain details of any other criminal activity, like burglary.

This begs the question: Why not? Obviously it’s easier to fingerprint and detect specific photos than to correctly detect conversations about “criminal activity” in general, but if they could do so reliably - should they? Is this the future we want - normal citizens lives being invaded by the police because their Gmail accounts showed “suspicious, possible criminal activity”? What level of false-positives would be acceptable?

Should such detection be reserved only for universally agreed morally-reprehensible crimes such a exploitation of children (vs say insider trading)? If so, who gets to categorize crimes as morally-reprehensible or not, Google?

Sounds scary to me. I can see why Gmail’s rep “made clear” that this isn’t currently what is happening.