Question 1

What is uncensored AI?

Accepted Answer

On LibertAI, uncensored AI means an assistant that answers to you rather than to a brand-safety department. There is no extra corporate filter layer stacked on top of the open-weight model, no surveillance or logging of your conversations, and no silent behavior changes. You control how the assistant responds through your own system prompt. It does not mean lawless: the models keep whatever safety training their creators built in, and terms of service still apply.

Question 2

Is uncensored AI legal?

Accepted Answer

Yes. Using open-weight models is legal, and so is running them without an extra corporate moderation layer. What stays illegal is illegal use: anything against the law is still against the law, and it is also prohibited by the terms of service. The point of uncensored AI here is removing surveillance and an opaque policy wrapper, not enabling misuse.

Question 3

Do the models have any built-in safety?

Accepted Answer

Yes. Every model keeps whatever safety training its creators included when they released the open weights. LibertAI does not add an extra moderation layer on top, and it does not strip anything out either. You get the model as released, so its built-in behavior is exactly what its creators shipped.

Question 4

Can I customize how the AI behaves?

Accepted Answer

Yes. chat.libertai.io supports custom personas, and you can write your own system prompt to define the assistant's tone, role, and rules. Because there is no hidden policy wrapper rewriting your instructions, the behavior you specify is the behavior you get, both in the chat and through the OpenAI-compatible API.

Question 5

Why does decentralization matter for censorship?

Accepted Answer

When AI runs on a single company's servers, that company can deprecate your workflow, pull a model, inject hidden instructions, or rewrite the rules overnight, and you have no say. LibertAI runs on Aleph Cloud's decentralized infrastructure with open-source code, so there is no single kill-switch and no single party that can quietly change what you are allowed to ask.