I wrote this reply below in response to CreamMeAgain:
https://umd.net/forums/i/ai-attitudes-survey-2025?start=60#postid_1083773
I decided to move this to a different thread because it does not have much to do with the poll. Here is my reply:


AI can be like piracy. It is not just the content that gets generated, but there are AI-models

Models need to be aware that their photos may be used to train an AI-model (or often called lora models). These AI-models are distributed and are often available on the generator websites. These AI-models are able to consistently reproduce the likeness of a person. If one person buys a photoset to get photos and train an AI-model on it, then others do not need to purchase the photoset if the AI-model is distributed. So this is exactly like piracy. Also, the AI-model allows you to generate nude, sexualized, or wammed images of their likeness.

Models now need to search for AI-models similar to how they search for piracy. I suggest using search engines and searching for their name plus "lora" or "ai model", or try searching on the popular AI websites.

I found 3 models that have made WAM content on sites hosting their AI-model:

Chloe Toy
Ashley Lane
Kendra James

Attached are images as proof with their faces blurred out to avoid breaking rules. I am assuming these AI-models were created without the Model's consent, and that they do not like their image being used to create nonconsensual content even if it is fan art.

I'm not even sure what Models can do to take them down. This is not like something which a DMCA takedown request can be used.

I knew about this 2 years ago, but I didn't want to say anything because I didn't want to spread the knowledge. Also, since then, it has become so easy with the tools because they require fewer images and even a single image. I hesitate to post this because I am making a lot of people aware that these AI-models exist, and the can use it to create non-consensual AI wam content or nude sexual content. People are figuring this out anyways.

It is also possible to reproduce content that simulates a producer's work similar to how they can reproduce an artists painting style. This is also similar to piracy. I will admit to training an AI-model, generating content, and sharing images on this forum a long while back. I never distributed the AI-model though. I stopped last year though. I will admit have generated nude wam content of celebrities for my own personal use, but I stopped doing that as well.

Here are some discussion points and questions:

Should discussions about how to create nonconsensual content be allowed?

With the rules right now, someone can post the links to these AI-models and a guide to how to create AI wam content of these Models or celebrities.

This would be difficult to moderate because I could discuss a generic workflow that can be adapted to make non-consensual content.

I think discussions should be disallowed when they specify names Models or celebrates. It should be against the rule to request nonconsensual content too.

Should links to nonconsensual content be allowed?

With the rules right now, someone can upload nonconsensual content to youtube or other video platforms, and I can post links to them. Someone could even setup a store website or patreon to sell it and link to the nonconsensual ai wam scenes here as long as they don't upload content.

This would be difficult to moderate too because it is verify difficult to be aware of every celebrity or person. This seems similar to the other rules about linking because it is as difficult as guessing the age or target audience of videos. I would guess it would rely on reports to identify content that breaks the rule.

I think the links should be disallowed. I don't want to see this site get a reputation for being a community that shares nonconsensual content. Links to piracy are not allowed. Links to minors content is not allowed. Links to nonconsensual wam content should not be allowed.

Is there anything we can do to help these Models?

I have no idea how to contact these Models. Do they know about this? We have a piracy report tool, so I am wondering if something similar is needed for future cases involving members here. Or maybe we can use the piracy report tool.

It is crazy how anything can be created by anyone that victimizes anyone, and it is becoming incredible easy to do so.

If Disney, Universal, Warner Bros, etc..can't do anything to stop AI platforms, certainly an individual's lawsuit is going nowhere.

This is an important post.

This is completely and utterly wrong on a moral level and whoever did these would be committing a sexual offence if they lived where I live.

These are porn models who have done some WAM, rather than models who only do WAM. I would assume whoever made these is not a wammer as the example images are not wam images. I'm not saying this makes it any better, but this most likely isn't one of us, at least.

You are correct regarding teaching "non-consensual methods", I want to make it very clear to people who don't know the technical stuff: there is absolutely no way of teaching someone how to train a Lora legitimately for an AI model without it being self-evidently obvious that the same method could be used to make non-consensual images. A good analogy would be like how it's impossible to teach someone to use a rifle for hunting without it being obvious that they could use that rifle in the same way for murder. I'm sure a few familiar names will argue that that means we should ban any discussion of AI techniques, but I thought that most of you lived in the land of the free, and I would argue that the individual is solely responsible for their actions and how they use their knowledge.

I strongly agree that links to non-consensual material should not be allowed. Again, depending on the exact content, where I live it it could be considered a sex crime. On a moral level as fetishists, allowing it to to be linked to from this site would be similar to allowing material featuring kids to be linked to from this site. I know MM is considering these things, and I think this should be the next policy he changes.

I think posting non-consensual R-rated or X-rated images should be result in being banned, whether AI or non-AI. There's absolutely no excuse for it.

However, proving it is non-consensual another matter

MM makes it very clear in the TOS that users are responsible for their actions and the consequences of their actions. It would be good if MM made clear a protocol for reporting and 'calling out' people that was fair to everyone involved. It would be terrible for someone to get falsely accused of this over a petty feud and to get caught up in a witch-hunt, but it would be just as bad if offenders just got their photos deleted but nothing else happened.

I'm not sure fetish models will have a lot to worry about when AI gets to a stage when people into WAM will be able to prompt; "Ok, a girl who looks like Madison Beer and one who likes like Olivia Rodrigo have a pie fight, then fall into a mud pit."

I'm not exactly sure what you're trying to say about whether this is a good thing or a bad thing, but the scene you've described above is already possible to make with AI. It would involve making and editing together a series of 5 second shots, using photos of Madison Beer and Olivia Rodrigo (with their consent, to be moral and legal about it) to train Loras. You'd probably also need to train a Lora on existing WAM videos or clips from the Great Race for the pie fight segment as well (again, with consent). It would take some effort, but you could do it.

What we're discussing is people doing exactly the above without consent, and fetish models will have plenty to worry about.

I will answer these separately but wanted to reinforce what mFeelz posted first.

Firstly, I have it from the highest authority and her own mouth that Kendra absolutely does NOT consent to her images being used in that manner. You will find that feeling is shared across the board among many of the models out there.

1) It's a violation. They did not give you consent to use their images in whatever AI fantasy you want to create and then share because you are too cheap to reach out to them and buy a custom. It is their body and their face so they will decide how, when and where it is shown, used, and displayed.

2) It is illegal. Never mind the moral issue of the fact that you are essentially violating a woman's autonomy, it is their likeness.

-You didn't hire them for the purpose.
-You didn't request consent for the purpose.
-You were never granted permission for the purpose.

It would be my suggestion to all models to have an attorney write up your own use of likeness paperwork that specifically forbids producers and photographers using your image to train AI. If they refuse to sign it, walk. This may not stop what is possible with what images are available on Google but it will give you legal recourse if you need it.

Yeah I think fetish models and producers do need to worry about this. It is true that there is probably a high demand for celebrity wam, but there are also a lot of fans of fetish models. If there was an AI-model or lora available, maybe potential custom customers would choose to try and generate what they are looking for rather than asking for a custom. It is getting easier, there is a larger variety of tools available, and the quality of the output is getting better.

A lot of people are into the common wam tropes like getting pied or slimed with clothing. It is becoming easy to make this content with any person, celebrity, and fetish models just by using some existing photos. I think you are right though that a lot of specific wam tropes and niches will require training.

No. The same applies to other prohibited topics here. We don't want the trouble it brings in addition to the fact that we don't want to encourage that behavior either.

It's been stated many times, but it's mind boggling to me that someone will post on this forum "Where are all the women at?" and in the same breath go "I aM jUSt mAkInG StUFf oN ai fOr MeEEee"

It's this attitude which is why the prevalent impression among women is we want to hurt them. Recently, AJ Lee's image was used by a fan who used an AI app to make them seemingly share a kiss and shared it online. Of course she saw this, expressed her reservation and asked for them to delete the post. He, of course, ignored her and it is still up.


No. Same explanation as above. Additionally, we don't need (or want) to have the same types who would consume content knowing it was created unethically.


Many women in the industry are well aware of AI and the filthy tactics being used. They talk about it frequently enough which is why I suggest women have lawyers draft up a "Use of Likeness" for their own protection of their brand and identity but at the end of the day, it boils down to what we permit. If we permit AI to be used with impunity of the repercussions, it's only going to get worse. At the same time, creative freedoms also need protection.

Flagging instances where their images or models are being used and/or shared so they can issue notices is a good start.

I agree with your sentiment, but do you think discussions on how to create consensual content should be allowed? Because it's essentially the same discussion. The techniques are universal, the subject matter is a personal choice. Grok already exists. Is "try Grok, it is less restrictive than the others" a discussion about how to to create nonconsensual content?


Who? I've seen both sentiments expressed but I've not seen both from the same people.


This is exactly the sort of non-consensual content we are likely to see here. Not explicit porn but creepy fan stuff. e.g. A fully clothed Taylor Swift pouring custard on a fully clothed Angelina Jolie.

I agree with everything else you've written.

Yeah you are right. There is nuance to this. For example, if we had a generic rule that prohibits discussion on how to create nonconsensual AI wam, then consider these following posts:

1) "How do I make AI wam of SomeSpecificPerson"
2) "How do I make AI wam of my wife who permission"
3) "How do I make AI wam using someone's photo"

This would result in the following:

1) It is clear this post should be deleted.
2) It is clear this post should be allowed.
3) This is hard to tell what the user meant. If they meant "a photo of someone who gave me permission" it should be allowed. If it meant "a photo a community user here that did not give permission" then it should be deleted.

All 3 discussions for my examples affectively have the same answer. Someone could give the same response to each one and tell the user what tool to use and workflows. It could become verify difficult to read a post and make a judgement to tell of the poster is trying to create consensual vs nonconsensual content.

For example, consider this post:
https://umd.net/forums/templates-for-ai

Or this:
https://umd.net/forums/ai-is-awesome-for-creating-pie-vid

For those two examples, how would a rule impact whether or not those posts are allowed? It is hard to judge.


I think something needs to change. Right now with the current rule, I think someone can:

* Make an AI-model (lora) using a community member's private photos, a producer's Model, or any person.
* Upload the AI-model to a generator website to distribute it.
* Share the link to the nonconsensual AI-model here.
* Discuss how they did it so that others can do the same.
* Ask others on the forum to create nonconsensual AI-models or content for them.
* Upload youtube videos of the nonconsensual AI content they created and share the link here.
* Create a store or patreon to sell the AI-model or content generated from it.

The only thing they cannot do is:

* Upload images or videos of nonconsensual AI content to umd.


I think we should consider the following rule:

1) Discussions on how to create content must not victimize an individual.

Examples:

* "How do I create AI wam video of Angalina Jolie?" - not allowed
* "How do I create AI wam from yearbook photos" - allowed

So someone would need to mention a specific person's name for it to break the rule. These posts or replies should be removed.

2) Links to AI content must not victimize an individual.

Examples:

* A link to the AI-model of Chloe Toy. - not allowed
* A link to a guide to train models or workflow that may result in nonconsensual content. - allowed

So if the link is targeting an individual, then the link should be treated like a piracy link and be removed.

My idea not perfect, but a step in the right direction.

Yes, I think this is the right sort of thing. It's the correct idea.

I am thinking that explicit guidance on how to break the rules shouldn't be allowed, but that discussion and guidance on AI training should be allowed, and this matches my own past ethical decision making. I've had to ask people not to ask me to make stuff for them on my profile, because I'd get a lot of requests about a year ago - "can you make a video of someone I know" - and I'd answer no. "Can you tell me how to?" and I'd answer no. I've never made a custom AI image or video for anyone, and I never will. However I have given advice on getting diffusion pipe running and choosing a container image on runpod.

Any rules would need to be set at a broad level using common everyday language, and not be technical. There is a lot of anti-AI sentiment from some users who aren't necessarily familiar with the technical side of it. Aside from comments from people who think this will be "inevitable in the future" without realising it's possible now, there will be people who'll think that non-consensual imagery is all you can use Lora training for, or that it's the only thing anyone would want to use it for (the actual thing that wammers want to use it for is to train concepts - the concept of a "thrown pie", the concept of "getting slimed", the concept of a "gunk dunk", and so on) .

Add to this that there are people who use terms like "training", "fine-tuning", "generating", "face-swap" and "deepfake" interchangeably (it's interesting that deepfakes have been around for at least 5 years now, and that this discussion hasn't really come up here until recently). If the rules get too bogged down in the nitty-gritty technical stuff, they won't be enforceable. People won't understand what they should be reporting, and moderators won't understand what should be enforced.

MM was very sensible at the start by stating "it doesn't matter about the actual technology used" when setting up the AI WAM group, and I think any moderation rules would have to follow the same sort of logic. My idea is that any rules about AI advice should also be applicable to advice about so-called "real" or "natural" material. In other words, rules about advice should be "AI neutral" - it should be the same rules for everything. In my mind, prohibiting people from asking about creating Loras because they could be used immorally, is akin to prohibiting people from asking about pie recipes because that pie could be used to ambush someone.

I've got a feeling we've almost got all the rules we need, we just need a bit of clarification on specific scenarios. Would love MM to weigh in on this when he has time.

I know exactly the sites you mean but whilst it's an important topic I think it's also important to separate out the true facts. This topic spans much more than wam content and is something that impacting the full industry.
CIvitAI, one of the larger hosts of open models has tried to address this after pressure from payment processors, They removed all character Loras of celebrities a while back and have taken further steps to prevent them but it's not perfect. Other sites have also followed similar paths to remove content. Now I'm no big fan of CivitAI but they are trying to remove this content even if it's due to outside pressures. SeaArt which is where you've picked up these links have not currently taken a stance as far as I can see. Their business model amongst others involves pulling and aggregating Loras from other sources without permission. My own wam Loras from last years were pulled from CivitAI and uploaded to SeaArt. It's one of many reasons I stopped uploading Lora's to CivitAI or elsewhere.

From a practical perspective, there isn't anything we can do to prevent someone either creating a character Lora or using an edit model to deepfake an image. The models that you highlighted here are 3 from 100's that exist either on that site or others. What we can control or rather UMD admins can control is what acceptable to post and link to.

Should discussions about how to create nonconsensual content be allowed?

As you say, there is no difference between the process of creating a completely AI character Lora to creating a celebrity or indeed anyone else. I'd argue the process isn't as simple as you've highlighted but the fact remains.
Edit models are more problematic particularly with Nanobanana pro releasing yesterday. Anyone "Could" upload a picture and edit it.

I think we can have a rule about AI that is absolutely neutral. Any time something happens during the production of WAM content without the consent of all parties it is an immediate ban. That would cover situations like Leon, and it would cover the situations where people are creating AI content of people without their consent.

Personally I agree with banning all non-consensual content, whether hosted here or elsewhere, though ultimately that's MM's call to make.

The problem is always going to be enforcement.

If someone creates fake WAM of a famous celebrity enough people will recognise and report it, kablam, gone. BUT, if someone goes to a mall, or a beach, or a nightclub, surreptitiously films or photographs some random person they find attractive, and then makes AI WAM of that person, and then posts it saying "based on my g/f who posed so I could use her as seed images, because she doesn't want to get messy but doesn't mind me AI messing her, as it avoids all the clean-up work" - how would any of us here ever know? Not everyone is going to be as obvious as to say exactly what they did, esp as people get more familiar with using AI to create their fantasies.

But I do think we should have the following as a rule, so at least people know, if they do otherwise, that they're breaking the rules even if they never get caught:

No non-consensual content at all, regardless of whether hosted here or elsewhere.

There's also been a lot of talk about protecting models and producers. The best way to do that, is to buy their content. Obvs only stuff any given person likes, but buy, don't just browse the sample images or download the trailers.

I look at it a bit like I look at piracy - our customers, for whom we are very thankful, are people who value our content enough to be willing to pay for it. They do that because they appreciate the time and effort that goes into creating it, and are willing to support us. Not because we make it hard to pirate, or threaten legal consequences, hexes, or other retribiution, but because they want to. Same reason I pay for movies instead of pirating them and buy music as downloads or sometmes even CDs - because I want to support the artists who created them.