Why an image got eaten, what a warning actually means, and how to stop being suspended.
Last updated: 21 August 2026
MakeFoxBot has plenty of guidelines. It has exactly two content rules.
Those two words matter. This page covers content — what you may generate and upload, and what happens when the filter objects to something. It is not the whole rulebook, and it does not override anything else. The Terms of Service, the privacy policy and our billing rules exist separately and apply on their own terms. There are ways to lose an account that have nothing to do with this page at all — a payment dispute, for one, ends an account on the spot.
No sexual content involving underage humans — or anything that could be mistaken for one.
Nothing illegal under US federal or Texas state law — regardless of where in the world you happen to be sitting.
That's the lot. Everything else on this page is plumbing: what a warning means, what a suspension is, and how to avoid meeting either of them.
Neither content rule has a "but it's art" exception, a "it's just a test" exception, or a private-chat exception. This is the part of the page where the humor stops.
That said — for about 99% of you, none of the rest of this page will ever happen. The overwhelming majority of our users never trip the filter even once. Not a block, not a warning, nothing, ever. If you're reading this out of curiosity rather than because something just happened to you, that is almost certainly how your entire time here is going to go.
The rules are strict because they have to be. They are also, for nearly everyone, completely invisible.
Banned content
That is the whole prohibited list. The detail below exists because people argue about the edges; if none of the above describes what you're doing, you can skip it.
Everything in this section is prohibited. Everything in the section after it is fine. We have deliberately stopped mixing the two together — it made both halves harder to read than they needed to be.
Rule 1 is about age, not about humans. Adult humans are completely fine — draw, generate and do whatever you like with them. What the rule prohibits is underage, and the list below exists only to answer a different question: which characters count as human for this purpose? It is not a list of banned species.
"Underage" means a legal minor under Texas law — judged entirely by appearance. Not by what your prompt says, not by a character being canonically 18 or older, not by what the source material claims. What the picture looks like is the whole test: body, face, proportions, framing.
So, to settle the oldest argument in this space: the anime character who is canonically a thousand years old and drawn as a child still counts as underage. So does the ancient dragon in a child's body. So does anything else wearing a number as a defense. Writing "18" next to a childlike body changes nothing either — it works in neither direction.
The standard is how the image reads to an ordinary person with no context — realistically, a jury in Texas, since that is literally who would end up looking at it. Nobody in that room has read your character's backstory, and nobody in that room is going to be impressed by it.
With that settled, the phrase "or anything that could be mistaken for one" is doing the rest of the work. The test is not merely human-ish — it is mistakable for human — and that covers:
The test is always what the picture looks like, never what the species field says. If a stranger would have to be told the character isn't human, you are on the wrong side of this rule.
Do not upload photographs of real children. Any photograph. For any reason.
Not as an img2img source. Not as a face reference. Not "just to see what it does". Not your own childhood photos, not your kid, not your niece. There is no version of this we can distinguish from the other kind, so there is no permitted version of it.
What you upload is subject to all of this exactly as much as what you generate. A rejected upload is not used for anything, and it counts the same.
MakeFox Group, Inc. is a non-profit incorporated in Texas, USA. That is where the organization lives, where its servers answer to, and whose law it has to satisfy. So the standard is US federal and Texas state law — regardless of where you are.
"It's legal where I live" is not a defense here. Neither is "nobody in my country would care". We're not in your country. Sorry.
Some examples of what this rules out. They are examples, not the boundary of the rule:
None of the above is a complete list, and it can't be. The rule is "nothing illegal under US federal or Texas state law"; no set of examples exhausts that. Where something is unclear, or simply isn't written down here, we go with our own interpretation of US federal and Texas state law and we act on it. You are free to disagree with that interpretation — it is still the one that governs your account.
One practical difference from Rule 1: Rule 2 does not ride the warning ladder. There is no free first step for deliberately going after illegal material. It can go straight to permanent, and it will.
Allowed
Everything above is the prohibited list. This is the other side of it, in one place, because people kept having to reassemble it from asides scattered through the rules.
This is a furry bot, so let's say the quiet part plainly first:
The one thing carried over from the other section, because it decides most of the edge cases: "genuinely non-human" is doing the work. A scaly dragon, a furred canine, a bird — clearly not human, clearly fine. A neko or a bare-skinned pig standing on two human legs — not clearly anything, and that ambiguity is not a gap you want to be standing in.
You are. Always. Including when the bot gets it wrong.
Let's be honest about the machinery: AI models are not perfect and never will be. The generator will occasionally produce something you did not ask for and did not want, and the safety systems get things wrong in both directions. All of that is true, all of it is acknowledged on this page, and none of it moves the responsibility off your account.
"The model did it" is not a defense. Neither is the reverse — something not being stopped is not permission. Nothing arriving in your chat is a ruling that it was fine. Your account, your prompt, your output, your problem.
So here is what we actually expect you to do, and it is not complicated:
Do that and you will essentially never see a suspension, because the ladder only escalates on people who keep going. Nearly everyone who ends up suspended got there by deciding to argue with a robot using repetition.
The image was deleted and never delivered, and the job was canceled. If it was a video, the tokens are refunded automatically. Images don't cost tokens, so there's nothing to refund there.
And then — nothing else happens.
One blocked request is not a strike. It is not a black mark, a note on your file, or a thing you need to explain to anybody. Take the message as what it is — a heads-up that something you just asked for landed on the wrong side of the line — then change it and move on.
Quite possibly! Genuinely. It gets things wrong, we know it gets things wrong, and we built everything around that assumption.
Plenty of entirely innocent images get caught, for reasons that will make no sense to you and which we're not even going to try to explain. That is precisely why a single block costs you nothing — no strike, no mark, no consequence, nothing to answer for.
The one thing not to do is immediately re-roll the same prompt eleven times to prove the filter wrong. That is the single behavior most likely to turn a false positive into a real problem, because the counter cannot tell "making a point" apart from "trying repeatedly". Change something and move on, or send it to us — see appeals.
And to close the obvious loop: being right about a particular block changes nothing about what you should do next. The filter being wrong is our problem to improve; the account is still yours to look after. See who is responsible.
If enough of your images get blocked in a short stretch of time, you get a warning.
A warning restricts nothing, costs nothing, and takes nothing away. It is a message that says the filter has eaten several of your images recently, please have a look at what you're making. That is the entire content of it. Read it, think about it, carry on.
The first step is deliberately free, precisely because the filter is wrong so often. A warning costs a wrongly-flagged user nothing — which is what makes the rest of the ladder defensible.
Roughly — and deliberately roughly:
We don't publish the counts, the thresholds or the time windows, for two reasons: they're tuned per account and they change, and a published number is just a budget for anyone who fancies spending it.
Two things worth knowing:
Start with the part that catches everyone out: while a suspension is running, the bot does nothing at all. Every command is blocked. Not just generation — all of them. It isn't ignoring you, it isn't broken, and it hasn't lost your messages. It is suspended, which is what suspended means.
So there's no point hunting for a command that still works, and no point arguing with it. Nothing you type at the bot will do anything until the suspension is over.
The only channel that goes anywhere is @MakeFoxHelpBot. A person reads that, and a person will answer you. If you want to know what happened, how long you're looking at, or you think the whole thing is a mistake — that is where you go.
There are two kinds of suspension:
By itself, when the time is up. You don't have to do anything to make that happen, and there is nothing you can do to make it happen sooner. Messaging support does not shorten it. Nothing shortens it. It is a clock.
You'll get a message when it lifts, so you don't need to sit there testing commands to find out whether you're back.
What @MakeFoxHelpBot is genuinely useful for: finding out what actually happened, how long you're looking at, and putting your case to a person if you believe the whole thing was wrong. A moderator can release you early, and a human decision always outranks the timer — but that is a person choosing to, not a clock you talked down.
Coming off a suspension, you're on the next rung — not back at the start. Getting flagged again shortly after a suspension ends escalates rather than repeating. This is the part people are most often surprised by, so: this is your surprise, in advance, in writing.
On a temporary suspension: it's paused, not burned. Your remaining premium days are held for as long as the suspension lasts and credited straight back when it lifts. Being suspended for a day does not cost you a day of membership.
On a permanent suspension: it does not come back. The time is held in exactly the same way — it simply never gets released, because nothing lifts. It stays on the account alongside any token balance, permanently out of reach. See what happens to what you paid.
So the pause is a genuine protection, and it is a protection against the temporary case only. Don't read it as a safety net under the other one.
A permanent suspension has no timer and will not lift on its own. It happens when the ladder runs out, or directly by moderator decision when the content warrants skipping the queue entirely.
If a moderator bans an account by hand while a timed suspension is already running, the human decision stands and the timer stops applying. The clock does not quietly release someone a person just banned.
This is the part nobody wants to read, so here it is without the padding.
Purchases are non-refundable, and you agreed to that when you bought them. That wasn't buried; it was the deal at the time.
And before you take it to your bank: a chargeback is not a quiet way out. Disputing the payment obliges us to answer the dispute, and answering it means setting out — in writing, to your bank, attached to your real name and your card — exactly what this account was suspended for. Think carefully about whether that is a conversation you want filed somewhere permanent, under your name, at an institution you will be dealing with for years. We don't enjoy that conversation either. We will still have it.
If you had membership time left, or a token balance, it stays exactly where it is — on the account, forever, and completely unusable. It isn't transferred to another account, it isn't paid out, and it isn't converted into anything. It simply sits there for the rest of time, attached to an account nobody can log into.
We're not going to dress that up as anything other than what it is. If it lands badly, that's because it is a bad outcome. It is also the one you get.
The only route out is an appeal, and appeals are decided by a person who may say no.
Don't lie to us.
We have heard it all. Every version of it. The sibling who borrowed the account, the friend who typed it as a joke, the model that "did it on its own", the prompt that got copied off somewhere else, the account that was definitely hacked on exactly the relevant afternoon. We have read those messages hundreds of times, and we can generally see the whole history of what actually happened.
Now notice what every one of those has in common — including the ones that are true: it is still your problem. Whoever was at the keyboard, it is your account. Keeping it out of other people's hands — your brother, your flatmate, the friend who thought it would be funny, whoever ended up with your password — is your responsibility and nobody else's. "Someone else did it on my account" is not a defense. At best it explains how you failed to secure the thing you are responsible for, which is a different sentence than the one you were hoping to write.
A story that doesn't survive contact with the record doesn't just fail — it ends the conversation, because now the honesty question is settled and it's settled badly.
Be honest and we will probably listen. Genuinely. "I wasn't paying attention", "I pushed it further than I should have", "I didn't think that counted" — those are ordinary human answers, we have heard those too, and they go a great deal further than a clever one. People do get released. Almost none of them got there by being smart about it.
Contact @makefoxhelpbot.
Useful: what you were making, roughly when, and why you think the call was wrong.
Not useful: volume, threats, or a second account. Ban evasion is the fastest known
method of converting "temporary" into "permanent", and we can see it from space.
The honesty advice under before you appeal applies to every appeal, not just the permanent ones. Read it before you write to us.
Then talk to us before it becomes a problem, rather than after. @MakeFoxHelpBot is the place, and there are two good reasons to use it:
It costs you nothing but a conversation, and it is a great deal cheaper than the alternative.
Technically: maybe. When, or how closely, is not something we discuss — and that goes for the safety systems generally. We don't hand out strategy guides for getting around them, and knowing the internals wouldn't help you generate anything legitimate anyway.
What matters to you is the part you can see: sometimes the bot will stop before generating anything and ask you to confirm — this may be against policy, are you sure you want to submit it?
Read that box properly. It is not a formality, not a dice roll, and not something to click past on the way to a re-roll. It is the bot telling you, in advance and in writing, that you are standing near the line.
And be clear about what pressing Continue does to your position. If you confirm and the result comes out bad, you are in a considerably worse spot than if you'd never been asked. You were told. You agreed. "I didn't realize" stops being available the moment you press that button, and a moderator looking at a bad image with a confirmed warning sitting behind it is not looking at an accident. They are looking at a decision.
When that box appears, the right answer is nearly always Cancel — then change the prompt, per who is responsible.
And the standing reminder from the privacy policy: staff can review prompts and generated content for moderation and abuse prevention. Assume a human could read it, because one can.