AI-generated child sexual abuse images could flood the internet. A watchdog is calling for action
AI-generated child sexual abuse images could flood the internet. A watchdog is calling for action
I am sort of curious, bc I don't know: of all the types of sexual abuse that happens to children, ie being molested by family or acquaintances, being kidnapped by the creep in the van, being trafficked for prostitution, abuse in church, etc etc... in comparison to these cases, how many cases deal exclusively with producing imagery?
Next thing I'm curious about: if the internet becomes flooded with AI generated CP images, could that potentially reduce the demand for RL imagery? Wouldn't the demand-side be met? Is the concern normalization and inducing demand? Do we know there's any significant correlation between more people looking and more people actually abusing kids?
Which leads to the next part: I play violent video games and listen to violent aggressive music and have for many years now and I enjoy it a lot, and I've never done violence to anybody before, nor would I want to. Is persecuting someone for imagining/mentally roleplaying something that's cruel actually a form of social abuse in itself?
Props to anybody who asks hard questions btw, bc guaranteed there will be a lot of bullying on this topic. I'm not saying "I'm right and they're wrong", but there's a lot of nuance here and people here seem pretty quick to hand govt and police incredible powers for.. I dunno.. how much gain really? You'll never get rights back that you throw away. Never. They don't make 'em anymore these days.
How often does tracking child abuse imagery lead to preventing actual child abuse? Out of all the children who are abused each year, what percentage of their abusers are tracked via online imagery? Aren't a lot of these cases IRL/situationally based? That's what I'm trying to determine here. Is this even a good use of public resources and/or focus?
As for how you personally feel about the imagery, I believe that a lot of things humans do are gross, but I don't believe we should be arbitrarily creating laws to restrict things that others do that I find appalling.. unless there's a very good reason to. It's extremely dangerous to go flying too fast down that road, esp with anything related to "terror/security" or "for the children" we need to be especially careful. We don't need another case of "Well in hindsight, that [war on whatever] was a terrible idea and hurt lots and lots of people"
And let's be absolutely clear here: I 100% believe that people abusing children is fucked up, and the fact that I even need to add this disclaimer here should be a red flag about the dangers of how this issue is structured.
Of all the problems and challenges with this idea, this is probably the easiest to solve technologically. If we assume that AI-generated material is given the ok to be produced, the AI generators would need to (and easily can, and arguably already should) embed a watermark (visible or not) or digital signature. This would prevent actual photos from being presented as AI. It may be possible to remove these markers, but the reasons to do so are very limited in this scenario.
I respect your boldness to ask these questions, but I don't feel like I can adequately answer them. I wrote a 6 paragraph essay but using GPT-4 as a sensitivity reader, I don't think I can post it without some kind of miscommunication or unintentional hurt. Instead, I'll answer the questions directly by presenting non-authoritative alternate viewpoints.
The issue here is that it enables those who would make the actual CP to hide their work easier in the flood of generated content.
Animesque art is one thing, photorealistic is another. Neither actually harms an underaged person by existing, but photorealistic enables actual abusers to hide themselves easily. So IMO, photorealistic 'art' of this sort needs to be criminalized so that it can not be used as a mask for actual CP.
If their work gets flooded then wouldn't that decrease incentive to produce it?
Points about real stuff hiding in a sea of fake stuff aside, because these ais would likely have been trained on images of real children and potentially real abuse material, each new generated image could be considered a re-exploitation of that child.
Of course, i don't think that's true in a legal sense but definitely in an emotional and moral sense. I mean look at the damage deepfakes have done to the mentals for so many celebrities and other victims, then imagine literally a minor trying to move past one of the most traumatic things that could have happened to them
I really don't think it would actually be trained on that specific data to be able to create it. If it can figure out a blueberry dog "child naked" seems pretty boring.
Isnt it better the are AI generated than real? Pedophiles exist and wont go away and no one can control it. So best they watch AI images than real ones or worse
Yeah exactly, I don't want to see it but the same goes for a lot of weird fetishes.
As long as no one is getting hurt I don't really see the problem.
As long as no one is getting hurt I don’t really see the problem.
It'd be hard to actually meet that premise, though. People are getting hurt.
Child abuse imagery is used as both a currency within those circles to incentivize additional distribution, which means there is a demand for ongoing and new actual abuse of victims. Extending that financial/economic analogy, seeding that economy with liquidity, in a financial sense, might or might not incentivize the creation of new authentic child abuse imagery (that requires a child victim to create). That's not as clear, but what is clear is that it would reduce the transaction costs of distributing existing child abuse imagery, which is a form of re-victimizing those who have already been abused.
Child abuse imagery is also used as a grooming technique. Normalization of child sexual activity is how a lot of abusers persuade children to engage in sexual acts. Providing victimless "seed" material might still result in actual abuse happening down the line.
If the creation of AI-generated child abuse imagery begins to give actual abusers and users of real child abuse imagery cover, to where it becomes more difficult to investigate the crime or secure convictions against child rapists, then the proliferation of this technology would make it easier to victimize additional children without consequences.
I'm not sure what the latest research is on the extent to which viewing and consuming child porn would lead to harmful behavior down the line (on the one hand, maybe it's a less harmless outlet for unhealthy urges, but on the other hand, it may feed an addictive cycle that results in net additional harm to society).
I'm sure there are a lot of other considerations and social forces at play, too.
I guess it depends on what pedophilia is in the end of how it's developed.
If it's more like a sexual preference then it's probably there already when someone is born and not changeable, but if it's more like a fetish then those are (afaik) related to experiences and exposures while growing up and actually can change and develop over time - and in that case it could be really dangerous to have that kind of material floating around freely.
We absolutely can make pedos go away.
Now that CSAM content is generated by bigcos with deep pockets, politicians don't want to scan their servers or take any other action. These are the same demagogues who wanted to kill end-to-end encryption and scan ordinary people's devices in the name of CSAM. Greedy and hypocritical vermin.
Normally I err on the side of 'art' being separated from actual pictures/recordings of abuse. It falls under the "I don't like what you have to say, but I will defend your right to say it" idea.
Photorealistic images of CP? I think that crosses the line, and needs to be treated as if it was actual CP as it essentially enables real CP to proliferate.
I keep seeing people post this same idea, and I see no proof that it would actually happen.
Why would you need "real" CP if there's like-for-like-quality AI CP out there?
Also, aside from going out of our way to wreck the lives of individuals who look at the stuff, is there any actual concrete stats that say we're preventing any sort of significant number of RL child abuse by giving up rights to privacy or paying FBI agents to post CP online and entrap people? I Don't get behind the "if it theoretically helped one single child, I'd genocide a nation.." bs. I want to see what we've gained so far by these policies before I agree to giving govt more power by expanding them.
This is a difficult one to get morally 'right', but I can see how legal (or at least not-all-out-illegal) AI CP could make the situation worse. Given today's technological advances, it will be next to impossible for law enforcement to reliably distinguish between illegal real CP and not-illegal artificial CP, meaning images and videos of actual child abuse cannot be used as evidence in court anymore, as the defendant can just claim that it's AI-generated.
\
Second, while a lot of consumers of CP might be happy with AI material, I expect that for a substantial number, the real thing will be considered superior or a special treat... much as many consumers of 'normal' porn prefer amateur porn over mass-produced studio flicks.
\
The two combined would mean there's still a considerable market for real CP, but the prosecution of child abusers would be much, much harder.
Photorealistic images of CP? I think that crosses the line, and needs to be treated as if it was actual CP as it essentially enables real CP to proliferate.
While I absolutely don't want to sound like I'm defending the practice (because I'm not), I'm really not too sure of this. If this was true, would similar logic apply to other AI-generated depictions of illegal or morally reprehensible situations? Do photorealistic depictions of murder make it more likely that the people going out of their way to generate or find those pictures will murder someone or seek out pictures of real murder? Will depictions of rape lead to actual rape? If the answer to those or other similar questions is "no", then why is child porn different? If "yes", then should we declare all the other ones illegal as well?
It's not that I think AI-generated child porn should be accepted or let alone encouraged by any means, but as was pointed out it might actually even be counterproductive to ruin someone's life over AI-generated material in which there is factually no victim, as reprehensible as the material may be; just because something is disgusting to most of us doesn't mean it's a very good justification for making it illegal if there is no victim.
The reason why I'm not convinced of the argument is that a similar one has been used when eg. arguing for censorship of video games, with the claim that playing "murder simulators" which can look relatively realistic will make people (usually children) more likely to commit violent acts, and according to research that isn't the case.
I'd even be inclined to argue that being able to generate AI images of sexualized minors might even make it less likely for the person to move over to eg. searching for actual child porn or committing abuse as it's a relatively easier and safer way for them to satisfy an urge. I wouldn't be willing to bet on that though
Good thing it’s already outlawed by most places
The legality doesn't matter, what matters is that the sites will be flooded and could be taken down if they aren't able to moderate fast enough. The only long-term viable solution is image classification, but that's a tall ask to make from scratch.
It's already happening on Pixiv...
I can’t believe how hard it is to avoid drawn or generated cp on there— and you can only ignore one tag without premium, so it’s not viable to manually make a blocklist :(
That only-one-ignore-without-premium thing is really asshole design, though
🤖 I'm a bot that provides automatic summaries for articles: ::: spoiler Click here to see the summary NEW YORK (AP) — The already-alarming proliferation of child sexual abuse images on the internet could become much worse if something is not done to put controls on artificial intelligence tools that generate deepfake photos, a watchdog agency warned on Tuesday.
In a written report, the U.K.-based Internet Watch Foundation urges governments and technology providers to act quickly before a flood of AI-generated images of child sexual abuse overwhelms law enforcement investigators and vastly expands the pool of potential victims.
In a first-of-its-kind case in South Korea, a man was sentenced in September to 2 1/2 years in prison for using artificial intelligence to create 360 virtual child abuse images, according to the Busan District Court in the country’s southeast.
What IWF analysts found were abusers sharing tips and marveling about how easy it was to turn their home computers into factories for generating sexually explicit images of children of all ages.
While the IWF’s report is meant to flag a growing problem more than offer prescriptions, it urges governments to strengthen laws to make it easier to combat AI-generated abuse.
Users can still access unfiltered older versions of Stable Diffusion, however, which are “overwhelmingly the software of choice ... for people creating explicit content involving children,” said David Thiel, chief technologist of the Stanford Internet Observatory, another watchdog group studying the problem.
Saved 78% of original text. :::
I'm not brave enough at the moment to say it isn't some kind of crime, but creating such images (as opposed to spamming them everywhere, using them for blackmail, or whatever) doesn't seem to be a crime that involves any victims.
My bigger concern is the normalization of and exposure to those ideas and concepts (sexualization of children). That’s also why I dislike loli/shota media, despite it being fictional.
That said, I still think it’s a much better alternative to CSAM and especially to actually harming a child for those who have those desires due to trauma or mental illness. Though I’m not sure if easy, open access is entirely safe, either.
The same concern has been behind attempts to restrict/ban violent video games, and films before that, and books before that. Despite generations of trying, I don't think a causal link has ever been established.
For "normalisation of sexualisation of children" go ask the people organizing child beauty pageants.
I'm brave enough to say what I am sure some people are thinking.
If a pedophile can have access to a machine that generates endless child porn for them, completely cutting off the market for the "real thing", then maybe that's a step in a positive direction. Very far from perfect but better than the status quo.
The ideal ultimate solution is to develop a treatment that pedophiles can use to just stop being pedophiles entirely. I bet most pedophiles would jump on such a thing. But until that magical day maybe let's explore options that reduce the harm done to actually real children in the immediate term.
Some psychologists agree with you. Others say it would only make the problem worse, making them want to escalate. Definitely one that I'm letting the professionals debate on and I'll go with their opinion
Some can't be cured by their own admission, its seems for some it is a deep seated sexual deviation that was probably established in very eaely life.
Well, there's all the children whose photos were used for the training data. I'd consider them victims, since AIs can't produce truly new images, so real human victims were needed in order to make AI images possible. And it's been established that AIs need to be trained on new human-made content in order to develop, as the images become distorted when trained on AI-generated content, so unless the paedophiles can be convinced to be satisfied with the AIs as they currently are instead of wanting better/more varied child abuse images next year, a whole lot more real children will need to be abused and photographed in order to improve the AI.
There's a lot of misconceptions about AI image generators in here.
They can indeed generate "truly new images", ask an image generator for an image of something that definitely doesn't exist in its training set and it'll likely be able to come up with something like that for you. Most importantly for purposes of this discussion, you don't actually need to have any images specifically of child abuse in a model's training set in order to train it well enough to produce images of child abuse. Train a model with a bunch of regular porn and a bunch of ordinary images of children and I expect it'll figure out how to make images of children in sexual situations if you ask it to.
This has been known for years. These AIs are capable of "understanding" the things they're trained on and creating novel interpretations of those things.
There was an article recently that showed if you trained many generations of AIs on just the outputs of previous generations you got degraded performance over time, but that's a pretty specific scenario that doesn't match what's being done in real life. In real life synthetic training data (ie, AI-generated training data) can be very useful for expanding the capabilities of AI as long as it's well-curated (humans need to select good outputs and ensure they're described correctly) and ideally has some of the earlier training set's original data included as well.
Another worry could be: how do you know if it’s a real victim who needs help, or an AI generated image.
If we had access to the original model, we could give it the same seed and prompt and get the exact image back. Or, we could mandate techniques like statistical fingerprinting. Without the model though, it's proven to be mathematically impossible the better models get in the coming years - and what do you do if they take a real image, compress it into an embedding, then reassemble it?
Many "AI generated" images are actually very close to individual images from their training data so it's debatable how much difference there is between looking at a generated image and just looking at an image from its training data in some cases at least.
That's a symptom of overfitting, which requires the image to be repeated in the training hundreds or even thousands of times. That generally only happened in earlier image generation models, more "modern" ones ("modern" in this case being measured in months because this is such a fast-developing technology) have much better curation of their training sets to avoid exactly that sort of thing. Nobody wants AI image generators that replicate images from their training sets, what would be the point?
So if you want to find an image model that gives you a close duplicate of an existing image of child abuse, you'll need to find one that was sloppily trained with a training set that included hundreds of duplicates of child abuse imagery. I kind of doubt you'll be able to find one of those.