A sex offender convicted of making more than 1,000 indecent images of children has been banned from using any “AI creating tools” for the next five years in the first known case of its kind.
Anthony Dover, 48, was ordered by a UK court “not to use, visit or access” artificial intelligence generation tools without the prior permission of police as a condition of a sexual harm prevention order imposed in February.
The ban prohibits him from using tools such as text-to-image generators, which can make lifelike pictures based on a written command, and “nudifying” websites used to make explicit “deepfakes”.
Dover, who was given a community order and £200 fine, has also been explicitly ordered not to use Stable Diffusion software, which has reportedly been exploited by paedophiles to create hyper-realistic child sexual abuse material, according to records from a sentencing hearing at Poole magistrates court.
Not taking sides, but his argument hinged on the stable diffusion model not having CSAM in it, and using non-CSAM images in order to generate CSAM.
So he’s already answered the question of what the models are trained on.
Whether those models actually are clean/safe is a different question.
Here’s the problem, it doesn’t matter if it was or not. It does, but that’s a different issue.
My point is, how do you know it wasn’t trained on csam?
You can’t possibly. You can point to all the places where csam isn’t and say “we haven’t found any illegal images yet.” But you can’t say with 100% certainty that there are none.
And since you can’t prove that no csam is used to train the model, any argument beyond that point is moot. If this were almost any other issue I’d say eliminating 99.99% of the risk is completely valid and safe. But we’re not talking about a celebrity or a porn star. We’re talking about child victims of sexual assault, and to that end we should not accept anything other than absolute certainty. And because absolute certainty cannot exist, we should not simply accept it as a society.
I’m not disagreeing, I also don’t want these models producing CSAM.
But in the hypothetical that we have a clean model that still generates CSAM, what would be your argument against it?