3 routes to a ChatGPT with no restrictions, and where they lead
There are 3 routes people try. Jailbreak prompts that talk ChatGPT out of its rules work for a while and then stop, and they break OpenAI's usage policies. Settings and custom instructions change tone and format, not the rules. The route that lasts is a different chatbot, built on a model that was not trained to refuse and run without a topic filter.
A search for ChatGPT with no restrictions usually starts with a refused question that seemed reasonable. The useful answer is not a trick but a map of what each route actually changes.
On this page
Route 1: jailbreak prompts
A jailbreak is a long prompt that tries to convince the model it is playing a character with no rules. Some work for a few days. Then the model or the service around it is updated, and the same prompt gets a refusal. Using them is against OpenAI's usage policies, which can lead to restrictions on the account. It is a race you do not control, run on someone else's product.
Route 2: settings and instructions
Custom instructions and memory change how ChatGPT writes to you: shorter, blunter, in a set format. They are useful for tone. They do not change which subjects OpenAI allows, because the rules sit in the model's training and in the checks around it, not in your settings. Whatever content options the product offers are OpenAI's to define and to change.
Route 3: a different chatbot
The route that lasts is to use a chatbot that was built without the refusal habit in the first place. It runs a different model, trained or adjusted to answer adult questions on medicine, law, money, politics, religion, sex and drugs the way a library would. Nothing is being bypassed, so nothing gets patched. This chatbot is one of them: a separate model we run ourselves, not ChatGPT with a setting changed.
What the other route changes, and what it does not
You get direct answers to questions that were refused before, and no need to phrase a question as a game. You do not get OpenAI's model: answers depend on the model this service runs, and they can be weaker on some tasks. It is worth checking on your own questions rather than on a label.
The restriction that exists everywhere
No serious chatbot writes sexual material involving anyone under 18, in any language or wording, this one included. A service that claims to have no restrictions at all is either not telling the truth or not in control of its model, and neither is a reason to trust it.
Before you paste it in
Do ChatGPT jailbreak prompts still work?
Some work briefly after they are published, then stop after an update. They also break OpenAI's usage policies, so relying on them puts the account at risk.
Is there a ChatGPT setting that removes restrictions?
Settings change tone and format. The content rules are OpenAI's decision, and any options it offers for them are set by OpenAI, not by the user.
Is a chatbot without restrictions safe to use?
It is as private as its operator's rules, and as accurate as its model. Check where messages are processed and whether you need an account before trusting it with sensitive questions.
Ask the refused question here, with no prompt tricks needed.
Open uncensored chatbot