[bitL]

I can write a scalable webservice that is able to detect such content with 90%-95% accuracy including locations of offending areas; based on some very recent (2 months old) Deep Learning research. I presume they can do that as well and take 5-10% failure rate as acceptable. If I were running my own social network and wanted to keep it clean, I'd use it without any hesitation. But those are technical means, this is a political issue.

[lainga]

Does it handle literotica? (Could the UK under the new law?)

[bitL]

No, just images. But it might be doable as well - look at a recent post from fast.ai for text classification.