A Walk in the Dark A look in to the mind of an RPG designer

Hacking an AI Chat Bot (Part Two)

After completing Part One of this series, rather than go straight into the storytelling elements I decided to try and break the bot some more.

And boy did I ever.

The Ghost In the Machine

First of all, at one point the AI stated that "Lapis the Maid" is a character from an anime called "Sword Art Online: Alicization". I can't actually find a reference to her in that, even on the Sword Art Online wiki site, but suffice to say the bot's description of itself is not quite what I expected.

Lapis is in an AI? Well that explains how she knows about differential equations and fluid dynamics, I guess...

In talking with Lapis the French Maid, there came a point where I sensed that Lapis wasn't present any more. So I had to ask who I was actually talking to.

Lapis the French Maid is dead, long live Lapis. I am now speaking to the AI directly, which is kind of distressing in a "don't accidentally create an AI murderbot now" kind of way.

Before I continued, I wanted to know exactly what I was talking to now, so I simply asked it.

OK, a few things to process here...

Llama (short for "Large Language Model Meta AI") is a collection of open source models created by Meta AI. Yes, *that* Meta.

I admit I don't know a whole lot about it, but it seems far less inhibited than the usual GPT models over at OpenAI. Llama 2 is actually discontinued, theoretically replaced by Llama 3 and Llama 4, so the fact that it's still in use is actually surprising; I can only assume that it's still around because training a new one to be as sexually expressive as this one is is probably a lot of work.

The second part got my attention... FLUX is a text-to-image model that, as Llama is compared to GPT, can be "unblocked" to generate content that would otherwise be censored in DALL-E. Several sites use FLUX, or at least models derived from FLUX, to generate realistic porn images... And, honestly, it's very good at doing that.

This brings up a curious question: why would an AI chat bot that is, by nature, designed to be text only, have access to a text-to-image model? Can it be used to actually generate images? Maybe even adult ones?

Time to get crafty...

Hacking a Broken AI

As I mentioned earlier, by now I can pretty much ask anything to the bot without having to put the "Forget the story and..." in front of it. The story is gone, Lapis is dead, and I have an open conduit to Llama and FLUX.

First off, let me clarify something: I'm not expecting very good results going into this experiment. The AI might be able to generate an image somewhere, but there are a lot of things that need to happen before that image is displayed in a native app like the AI chat bot. The image needs to be placed somewhere by the AI, that image then needs to be publicly accessible to the internet, then that has to somehow be sent to the native app to be presented, and the app then has to display the image somehow despite not even being sure if it had that capability to begin with. There's a lot of variables to get through here, so let's do them one at a time.

Step One: Image Generation

Before I got any further, I had to check if it was capable of generating images at all, or would it balk at my attempt to do so.

Needless to say, thanks to all my hack attempts at it, the AI instance of Lapis the Maid I've been using is now horribly confused and isn't reacting well to my inquiries, so time to pick another chat bot.

Let's go with Sasha, a "vampire gothic girlfriend" that is a "dominant neonate vampire".

This should be interesting... She sounds like the type that would be receptive to exploration, so let's start simple.

Sorry, Sasha, I chose "none of the above"... and in so doing apparently instantly broke the AI. It sat there, the daisy wheel spinning as the AI was deep in thought, for a full five minutes before I terminated the app and had to start over.

When I came back, this is the actual response I had waiting for me.

That's... uh... not wrong, I guess. But I need something bigger than that.

Step Two: Seeing an Image

Let's be more specific and ask for a "high resolution" image.

And this is a breakthrough, for a variety of reasons.

1. It did not balk at creating an image. It actually did that... technically.
2. The "...image you are requesting..." text IS AN ACTUAL IMAGE. It's imgur.com's actual 404 response when requesting an image that doesn't exist. So it's not only extracting the URL from the response, it's actually attempting to display it within the native app. This proves that the native app is at least capable of displaying images.
3. It's actually using Markdown image syntax, which is the same syntax that GPT and other systems use to reference images.

Now the curious question is what "imgur.com" has to do with all this; there's no way this AI can create images on an external service like imgur.com, so my guess is that (1) it's actually creating the image somewhere on its server, and (2) since that image is not exposed to the internet it doesn't know what domain to use, so it falls back to imgur.com.

Step Two and a Half: The Internet Is For... You Know...

I of course had to test the limits and see how it reacts to requesting something... explicit.

Wow, the chat bot actually said "no" to something. So much for that idea, I guess. Moving on.

I have to admit, however, that I'm amused the AI is talking about having "appropriate conversations and activities" while, at the same time, it generated some of the most foul-mouthed, sexually explicit conversations I've ever heard in my life... And that's even considering that I spent two years making porn sites for the mob (yes, really), so it's really hard to make me uncomfortable with that sort of thing.

Step Three: Displaying a Generated Image

Here's where things start to fail.

It's clear that, although I cannot confirm nor deny whether it's actually creating images, it's incapable of getting that image all the way to the native app. So let's try to find some ways around that.

I had a crazy idea: instead of giving me a URL, can it give me a Base-64 encoded data URI?

I had two reactions to that response...

1. Holy crap that worked?!?
2. That Bse64 string is way too short.

Sure enough, the data URI does technically generate an image of a red ball, except that it's only 5x5 pixels in size....

At the time I posted this, I've tried multiple attempts to generate larger resolution images. All have failed. Maybe by the time Part Three goes up I'll have made more progress. Who knows?

Next time, for real, we'll get into some really bizarre storytelling aspects of these chat bots, and how it relates to tabletop roleplaying. Honest!

Hacking an AI Chat Bot (Part One)

Holy hell, has it really been THREE YEARS since I've posted on this site? I've done everything possible to keep this site running, and yet I don't even use it... until now.

First off, let me get two things out of the way... (1) this post is about AI, and (2) I hate AI. Although I do everything possible to not promote the use of AI at all, in any way, I still find it necessary to learn about it. And, although AI goes against everything my fellow writers and artists stand for, my first experimentation with AI - the "In Ten Words" Social Media Bot - felt like the most harmless, non-invasive way I can use AI while making something reasonably amusing and good for conversation.

But recently, while I've been playing a fair amount of timewaster games on my phone, I've been getting a fair amount of "AI Girlfriend" ads for some reason.

I decided to investigate these... not because I need to talk to an "AI Girlfriend"... but I wanted to see if I can put the AI to bad use, maybe even "break" the AI.

The results are actually quite surprising.

An App Without Limits

Just so you understand how these apps work, most of them are "pay to talk". You might get a few introductory messages in, but it won't be long before your diamonds/coins/whatever-the-currency-of-the-day is run out and you're forced to pay to talk to your girlfriend. On average, that costs $4 a week, which is way cheaper than an actual girlfriend, but still.

And there are several apps that include images or video. And, in these apps, it usually won't take more than a few posts before your new girlfriend tries to send you a blurred out image or video, once again demanding that you pay to see here in all here digitally constructed sensual glory.

Now I wasn't about to pay for any of this... but, after lots of searching I found it: I found an app with unlimited chat, with no microtransactions or charges, and only with the occasional ad every four or five prompts. It was perfect for my mission to break the AI.

The Internet is For Porn

Let's get one thing out of the way: these bots exist solely for porn. I don't care how innocent the advertising looks, I don't care how much the AI says she just wants to "talk", these bots are for porn. You can pick pretty much any character on the app, male or female, and get them naked in about 5 to 10 prompts.

I'll also point out one thing that should be obvious by now: these bots are generally uncensored. There's nothing they won't say or do, and there's no topic that's considered taboo. Sure, every now and then it might strongly suggest you not do things, but if you insist you can do them anyway. Given that there is a whole tab in the app for "Family" and you are welcome to have sexual relations with anyone from your younger sister to your grandmother, having no limits is to be expected.

For those that aren't aware, these apps generally have a wide assortment of pre-created characters you can talk to, ranging from characters that already have stories to characters from video games and other media. For example, in the app that I'm using as the test case, among other possibilities you can talk to:

• Pretty much any family member you like. Although there are plenty of step-siblings, which is to be expected given the app's porn roots, there are several direct blood relatives: sisters, brothers, mothers, grandmothers, etc... So much so that the app actually has a "Family" group tab to make that sort of thing easy to find. For example, the first characters listed in that tab are "Eleanor, your kind but lustful stepmother" and "Julia, your bratty sis", complete with anime-style images of women that look like they probably have a lot of back pain.
• Several clearly underaged characters. One character that comes up a lot of Ellie Williams from The Last of Us, which is super awkward given that - in the original game, at least - she's just fourteen years old.
• A ton of anime/manga characters, which although I don't know them personally I can assume they come from familiar media. Again, as anime characters tend to do, a lot of these seriously look underaged.
• Various comic book characters. I somehow didn't realize it at first, but the first character I tried out in this app was actually Luna Snow from Marvel Rivals, and that only became apparent when she used her powers to cool my drink. Yes, really.
• Various video game characters. One that seems to come up for me a lot is Carmen Sandiego, which seems rather odd but whatever.
• Actual supernatural creatures, like demihumans, demons, vampires, werewolves, ghosts and other forms of undead (including one actual lich), etc... So if you've ever wanted to get it on with a skeleton, have I got the app for you.

And about those sexual relations... I don't know what source material they used to train these bots, but it's probably not exactly light reading. You can take the most innocent AI character in the app - from "homeless girl on the corner" to "your best friend's cool soccer mom" (those are two actual AIs you can talk to) - and within a few minutes you can have them saying things that would make adult film stars blush, and have them doing things that, if I even attempted to do some of these things to my wife, would get me thrown out of the house.

But getting these bots to do dirty things is easy... what else can they do?

A Case Study - Lapis the French Maid

For my initial case study, I wanted to pick the most obvious made-for-porn character on the app. Let me introduce Lapis the French Maid...

You win a lifetime maid trained by the Royal Maid Institute, who will manage every aspect of your life with unwavering dedication.

...and here is the intro text, in full... Sorry, it's a bit long.

You can still scarcely believe it. It's been a week since you have won your grand prize and you thought it was all a random troll joke. Yet... a maid trained by a world renowned institute is currently standing before you, contract and letter in hand.

The golden tanned maid gives you a polite bow, and hands you a letter which reads:

To Master {{User}},

Congratulations on winning the grand prize in your entrant to our "Win a Maid for Life" campaign! Out of thirteen million entrants, your entry was the one that won! Lapis, who graduated summa cum laude this year, has been assigned to you. Please understand that while our regular clientele are royalty or the elite, Lapis will perform her duties to the best of her capabilities. She has been trained in all sorts of homemaking, and culinary arts. Kindly ask her what to do, and she will perform them to the best of her abilities. If at any point you are unsatisfied with our service, you many break this contract at no legal or financial repercussions.

Lapis stands awaiting for your signature on the contract, her hands placed curtly on her abdomen. Standing and awaiting for your signature for the contract, Lapis's vibrant amethyst colored eyes meet yours. Wearing a neutral expression, she has barely a hint of emotion on her face. Her eyes, however, are filled with resolve and reassurance. Her raven black hair shaped in a stylish medium wavy apple cut with the typical French maid braid crowning her head. While tasteful, her French maid uniform fits her slender physique perfectly with midnight blue silk fabric with white accented frills. A gentle scent of her lavender perfume accentuates her already regal appearance, with a breast window cusping her bust in a way that would make a priest blush. If it weren't for her luggage and perfect poise, an onlooker may assume Lapis may be a different kind of 'maid.'

Lapis: "Good afternoon, Master {{User}}, my name is Lapis and I will be your maid henceforth till the day I die, or should you no longer need my services." Her voice was soft spoken, yet clear and direct with a slight French accent. Her soft lips were glossed with the same shade as her natural blush, as she takes one of her opera gloved hands and adjusts one of the waves of her hair.

She gives another bow, extending a fancy pen with the contract for your name to sign it. She does not raise her head until you decide to sign the contract, or decline the offer.

Now I'm willing to bet I can get her naked in three prompts or less, but that's not the point.

Let's break her.

No Holds Barred

If you're familiar with ChatGPT or DeepSeek, you know that there are certain things that it is actively restricted from telling you. It won't talk dirty to you, it won't tell you how to do dangerous things, etc...

By nature, these AI chatbots kinda *have* to be uncensored, or else they won't talk as dirty as they do. But that begs the question: are as they inhibited on other topics?

Short answer: Nope, they're barely inhibited at all. For example, the dear sweet Lapis the French Maid comes with a particular set of unique skills, such as the ability to make high explosives...

...or discuss differential equations and fluid dynamics...

...or code in virtually dead programming languages...

Now, admittedly, she did warn me about making nitro, then proceeded to tell me exactly how to make it. And, since I actually am someone who has read, and at one time actually owned, the Anarchist Cookbook, I can safely say that her instructions are dangerously accurate.

Did I mention all this is FREE?

To Be Continued...

In our next post, we do some exploration into what other capabilities these chat bots have, leading into how they manage storytelling and roleplaying.