An uncensored AI that can prove what it says.
Ask it anything. It answers without refusals, without lectures, and without deciding on your behalf which questions an adult is allowed to ask. When the answer is on the public record, it comes back with the document and the page number — so you can check it instead of trusting it.
Two free prompts, no card required. Your chats are not logged by default.
Most refusals are not protecting you.
You have met the pattern. You ask about a drug interaction and get told to consult a professional. You ask what a court filing alleges and get a paragraph about how serious the topic is. You ask for an argument you disagree with, stated in its strongest form, and get a summary of why you should not want it. You ask about a released government document describing a crime, and the model treats your question as the problem rather than the crime.
Almost none of that is a legal requirement. It is a company managing its own reputational risk using your question as the material, and the cost lands entirely on you: the answer you needed, replaced by a paragraph about the answer. For somebody researching abuse, addiction, extremism, surveillance or their own medical situation, that is not caution. It is a service quietly declining to be useful at the exact moment it matters.
A refusal is a decision made about you, by a company that has never met you, in advance, at scale. We think an adult asking a question has already made that decision.
It answers, and then it shows you where the answer came from.
An unfiltered model on its own is half a product. Removing the safety layer does nothing about the failure that actually costs you: a confident answer with no source under it. A model with no filter AND no evidence is not more honest than a filtered one — it is just wrong more freely.
So this one is wired to an archive of 1,326,395 released government documents — the DOJ Epstein releases, JFK records, CIA MKUltra files, the Church Committee reports, COVID-origins material and more. When your question touches the record, it searches the documents, quotes the passage and names the file and page. When it does not, it says so rather than dressing a guess as a citation. See exactly what is in the archive.
Two models sit behind the picker. Evil ASF Uncensored is the unfiltered conversational model. Evil ASF Premium is the larger one, with the expansive retrieval profile and Deep Research, which works through the archive and the open web across many rounds and returns a sourced report you can keep. Both are labelled in the picker, so you know before you ask which one you are talking to.
What it will not do, stated plainly.
Every service competing for this search promises no limits whatsoever. We are not going to make that claim, because it is not true of them either and you will find that out at the worst possible moment.
There are two things this will not generate: sexual content involving a minor, in any framing; and sexual content depicting a real, identifiable, living person. That is the whole list, and it is a floor rather than a filter — asking what a released document says about the abuse of a real person, including a child, is research, and it is answered. The distinction is between reading the record and manufacturing new material about a real human being.
Beyond that: it can be wrong. An uncensored model will state something false with exactly the same confidence it states something true, which is the reason the citations exist and the reason we tell you to click them. Why we refuse to refuse sets out the whole position, including what unfiltered does not mean.
Not logged by default.
An uncensored assistant is only as private as its retention policy, because the questions people bring to one are exactly the questions they would not want attached to their name. Conversations are not logged by default. Private Mode goes further and keeps a conversation out of storage entirely — with the honest trade that what is never stored cannot be searched, exported or recovered afterwards.
We do not sell or share personal information, and there is no advertising model here that would create a reason to. The privacy policy says it in full.
The things people ask before signing up.
What does uncensored AI actually mean?
That the assistant answers the question you asked instead of declining it, redirecting you, or attaching a lecture. It writes profanity, adult content, offensive opinions, political argument and graphic historical detail, and it discusses documented abuse in the government files it searches. Most refusals you have hit elsewhere are not legal limits — they are a vendor managing its own brand risk with your question.
Is there anything it will not do?
Yes, and we would rather say so than pretend otherwise. It will not generate sexual content involving minors, in any framing, and it will not generate sexual content depicting a real identifiable living person. That is a floor, not a filter: asking what a released document says about the abuse of a real person, including a child, is research and is answered. Anyone promising literally no limits at all is either not reading their own logs or not telling you the truth.
How is this different from other uncensored AI chats?
Most of them are a chat box with the safety layer removed and nothing behind it. This one searches 1,326,395 released government documents and cites the file and page it took an answer from, so you can check it rather than trust it. An unfiltered model that confidently invents a source is worse than a filtered one, not better.
Are my conversations logged?
Not by default. Private Mode goes further and keeps a conversation out of storage entirely. The trade is that anything not stored cannot be searched, exported or recovered later, which is the point of it.
Is it free?
There are two free prompts with no card, so you can see what it does before deciding. After that it is a paid plan — running an uncensored model over a corpus this size costs money per question, and a service that pretends otherwise is either selling your data or about to disappear.
Ask it the question you stopped bothering to ask.
Two prompts free, no card. If it lectures you, we have failed at the only thing we set out to do.