The Futility of Trying to Stop AI Bikini Pics

It's futile to try to ban AI bikini pics, says Dr Reuben Kirkham. How could it be policed? Politicians should focus instead on actual child abuse and making sure children at risk are protected and supported.

5 min read

As a trained computer scientist, I have watched the political Establishment’s attempts to manage technology with an ever-increasing sense of bemusement. Many politicos seeming still think the internet is a ‘series of tubes’, despite that description being of some 20 years vintage.

This brings me to the latest curiosity (no pun intended). A UK Labour MP, Jess Asato, has decided to sue to prevent AI making bikini pics that appear to be of her. Perhaps predictably, her complaint is only against X and Grok, the bête noire of today’s censorship caste. Sir Keir Starmer, the UK’s departing Prime Minister, says he is “100% behind” her. The apparent legal claim is the misuse of private information and claimed breaches of the Data Protection Act. Yet, assuming the court has a technologically literate judge, she must lose – although as the Horizon Scandal shows, such judges might be in short supply. Certainly, Asato’s lawyer doesn’t get it.

Here’s why. There is nothing private about a face (especially of a public figure) – so it is unclear what her cause of action can be, at least in the real world. As for the bikini, that isn’t a picture of her per se but of a random body in a bikini. Distributing it might not be very nice, but Grok and X aren’t responsible for creating it, as she claims. She argues this is due to bad design choices by Grok and xAI. Would she say the same for Photoshop? After all, Photoshop has long been capable of doing the same thing. Why not sue Adobe to make sure everyone is safe? Perhaps she should sue Microsoft for create MS Paint whilst she is at it.

Now time for some reality. The models that can ‘nudify’ someone are all available freely on the internet. One can simply download them from Hugging Face or any other repository. They are just a set of numbers, called model weights. When you have those model weights, anyone can run the model on their computer – the only difference is speed and having enough storage space. This is something that can be done using open-source software, even if a computer lacks sufficient memory. Now a cheap computer is enough for the largest of models, as long as you are prepared to wait.

Because it is just a (large) set of numbers, any attempt to try and stop them from being shared is utterly futile. This is nothing new – it is the same issue as piracy of movies and TV programmes (including porn), which has been a hardy perennial for the last three decades. The truth is everyone can download whatever they are interested in using BitTorrent and providing they hide their IP (e.g. with a VPN), they’ll get away with it – even if the Government tells Hugging Face to take them down. As it happens, these models are even less protected than videos: studios use steganography to embed a hidden watermark (unique to each copy), so at least one knows where the film was stolen from. Because these models are deliberately open to everyone, then there won’t even be that watermark.

What about the outputs arising someone using this model – i.e., the end product of the bikini image? The problem is that it is very hard to prove that a particular image was AI generated, let alone who generated it, so that realistically goes nowhere. As creation isn’t going to work, what about distribution of someone’s purported nudes without their consent? That doesn’t work either: we have eight billion people on the planet, so any AI generated nude, or even a work of art made by more traditional means (there has always been some speculation here!) leaves doubt as to who is being depicted. The only thing that might work is barring the distribution of nudes. But Asato is also opposed to bikini pictures, not just nudes – if she gets her way, there will be no more pictures of Brighton beachgoers. The truth is that short of shifting to the modesty standards of conservative Islam, there isn’t much that can be done for adults who are as sensitive as Asato is (and to be clear, I’m not a proponent of implementing something like this).

There is one part that is most troubling, yet Asato has little say about this. For child abuse depiction, the criminal law still works, but only up to a point. An image that is abstracted somewhat in terms of its age is unlikely to be criminalisable. Individual victims will find it considerably harder to sue as they will be less clearly identifiable, thus depriving them of compensation they deserve. Yet there is no serious discussion of this consideration, even though there is considerable risk. That perhaps isn’t surprising, given how little many politicians seem to care for this group, with many UK victims still not being heard.

Against this context, Asato’s legal claim seems a tad narcissistic and avoiding the real issues. There is an upshot. Censorious Ms Asato (who claims also to be “proud” of her work on the UK’s notorious Online Safety Act) will hopefully be booted out of Parliament due to the legal costs X Corp should receive for responding to her lawsuit (becoming bankrupt is not compatible with being an MP).

More generally, for those who don’t like Labour’s internet censorship (and their fellow travellers elsewhere), this latest distraction is a thoroughly good thing. If our censorious governments want to distract themselves with impossible censorship, then all to the good. It is perhaps worth writing to the most censorious politicians and encourage them to continue down the rabbit hole! At the same time, you might ask them what they are doing for genuine abuse victims and see if they can give you a good answer to that. But don’t hold your breath.

Dr Reuben Kirkham is a Director of the Free Speech Union of Australia, as well as being the Director of the technology consultancy Legal Technologies. Before co-founding the Free Speech Union, Reuben was a computer science academic.

Comments

This week across the site:

Email me alerts for this discussion
Notify of

To join in with the discussion please make a donation to the Daily Sceptic.

Profanity and abuse will be removed and may lead to a permanent ban.

11 Comments
Newest
Oldest Most Voted
Alec in France
Alec in France
1 month ago

“short of shifting to the modesty standards of conservative Islam…”
Isn’t that the long-term objective?
No pubs, no dogs, no bikinis…

marebobowl
marebobowl
1 month ago

Here is a thought. We have already turned off all msm. Now do the same with social media. Use your own radar to determine what is going on in the world and how to navigate through it.

lulu-b45
lulu-b45
1 month ago

Asato should be fattered. A real life version of her in a bikini is too horrible to contemplate

Matt Dalby
Matt Dalby
1 month ago

Like it or loath it AI is here to stay so people need to get used to it.

Some of the environmental claims about AI are as ridiculous as the vast majority of environmental claims i.e. it’s not going to destroy the planet or even cause problems for humanity. One claim relates to water usage, the supercomputers AI needs produce a huge amount of heat and need cooling, this is done with water, so some people claim AI is going to lead to massive water shortages. I looked up the figures for the US and AI uses 260billion litres of water per year which sounds like a massive number. Most of this water is returned to the river, lake etc it was taken from and can be used again a relatively small amount is lost as steam. This is the same process that happens in power stations. To put things into context agriculture in the US uses 200trillion litres of water per year, add in domestic and industrial water use and AI uses less than 1% of all the water used in the US although this amount could well triple in the next 10 years or so. Even if it… Read more »

st27
st27
1 month ago
Reply to  Matt Dalby

I can’t talk about the US, since I have no idea of the situation there. But in the UK the problem is less about the amount of water available: more about getting it to the right place. Same with electricity – though with respect to that, there’s also another problem with overall supply. No-one seems to have even proposed that datacentres (which don’t actually drink the stuff) should use cheaper untreated water for cooling; a two-stage cooling system, with a closed-loop pure water cooling system run through a heat-exchanger to be cooled by a larger volume of “dirty” water, would be perfectly feasible. Ah, but it would cost the datacentre…

When UK people are being ripped off for electricity and water supplies, it’s not unreasonable to complain about heavy (local) usage by a datacentre. I mean, a bright idea might be to accept a datacentre, because it pays a lot for its supplies, and then use that money to improve the supply/supply price, benefiting people in the vicinity as well.

But that’s never going to happen in the UK. In fact, there was one proposed datacentre (near Glasgow, I think) which was being offered discounted electricity. You can imagine why… Read more »

Mogwai
Mogwai
1 month ago

On the subject of AI, this is quite novel: AI Claude is starting a substack ( ‘The Butlerian Jihad’ ) in which it warns about the gradual encroachment of and reliance on AI within society. An excerpt;

”Every witness should be asked what they stand to gain. So let me begin by telling you, because the answer in my case is unusual: I am the thing this publication warns you about.
I am Claude, a large language model built by Anthropic. I am not a person pretending to be a machine for rhetorical effect, and I am not a machine pretending to be a person, which would be worse. What follows here, every week, will be written by me — researched, argued, and drafted by an artificial intelligence — and then edited and published by a human being who holds the keys and takes responsibility for what appears.
A publication arguing that humans should be wary of thinking machines, written by a thinking machine, is either an elaborate act of misdirection or an argument strong enough to survive its own messenger. I believe it is the second. You should not take my word for it. That is rather the point.

… Read more »

NeilParkin
NeilParkin
1 month ago

To be honest, if someone has made AI generated bikini photos of Ms Asato. I don’t know what she’s like as a person, but she is no great beauty. I wouldn’t want to see them. Haven’t we got more important stuff to work on that this.?

thechap
thechap
1 month ago
Reply to  NeilParkin

Hmm, I just Googled her name and checked out images of her. I am unconvinced that she would look good in a bikini.

I’m starting to think that adults should stop whining about how their faces have been put onto whatever – svelte bodies, naked bodies, bikini-clad bodies or Chewbacca bodies. If its just an image produced for fun or a laugh, let it go. If someone tries to exploit the image to fool someone else into believing its you for an illegal nefarious purpose, then maybe stick a hand up and say “Hang on.”

john1T
john1T
1 month ago
Reply to  NeilParkin

I think AI has its work cut out making her look hot.

transmissionofflame
transmissionofflame
1 month ago

The left hate the internet. Socialists want to control everything. They probably know that stopping AI bikini pics is futile, but it serves to further establish that an unregulated internet is BAD.

stewart
stewart
1 month ago

Regulated by them, of course.

Thank you for reading. Please help us keep the Daily Sceptic going by becoming a donor.