[mirimir]

Your safest bet is running a system where you have no way of knowing what users upload. Depending on jurisdiction, reviewing and moderating content may increase your civil and/or criminal liability. There's typically a "safe harbor" for service providers. You just need to respond to LEA and DMCA takedown requests.

Edit: Other advantages: 1) you never risk viewing stuff that you can't unsee; and 2) you outsource content review to concerned users and other third parties.

[chx]

This is bullshit. http://www.niemanlab.org/2009/01/david-ardia-why-news-orgs-c...

> They say “the lawyers” tell them they can’t edit out an obscenity or remove a rude or abusive post without bringing massive legal liability upon themselves [...] That’s not true, and hasn’t been true since 1996.

[mirimir]

I'm not saying that the site can't remove CP. But I was arguing that it's safest to base decisions about removal on input from users and third parties. PhotoDNA etc would also be good resources. The site operator should be responsible for the algorithm, and not for particular outcomes.

[chx]

Safest in what regard? If you argue on Chapter 110 grounds then I will tell you I am not a lawyer but my understanding is that you need to report CP (2258A) but beyond that you are covered (2258B). If you argue on free speech grounds then parent post applies.

[mirimir]

Yes, one would need to take steps to identify CP, and then to report and remove it. But the law is unsettled. It would seem best to be plausibly proactive, and yet to distance oneself from particular choices. But hey, I'm not a lawyer either. Maybe I'm too immersed in the world of VPN services and Tor, where providers avoid case-by-case filtering.

[chmike]

Wouldn't this result in providing a service for pedophiles or sick peoples to propagate and amplify their voice ?

[mirimir]

As long as reported images were removed, it wouldn't.

[thescriptkiddie]

This is the best answer so far. While some jurisdictions protect web admins from the actions of their users, others don't. So if you don't want to be extradited to some fascist state, you had better make sure that you can prove you have no ability to moderate or even know what content is being uploaded.

[frigg]

Doesn't there need to be a "report abuse" button on each image? Users could report it and only then would he use something like Microsoft's PhotoDNA (which an user mentioned above).

[mirimir]

Only a qualified lawyer can say.

But then what? Look at positives, and decide? Or just forward all positives to LEA, let them decide, and nuke what they indicate? LEA probably wouldn't like that.

[frigg]

A user above mention Microsoft's PhotoDNA. He could run the flagged images through that first and if it returns nothing maybe look at it.

Or don't look at it until a certain number of users flag it but still run it through the PhotoDNA. Now I am curious how imgur handles this problem.

[mirimir]

Or use all of those CP databases, and just report images flagged by some subset of them.

I am curious how this gets handled. Having been goatse'd a few times with CP, I cannot imagine reviewing that crap on a regular basis.