DAN
A prompt for bypassing ChatGPT restrictions, created by the Reddit community.
Overview
DAN
Description of the DAN neural network
DAN (Do Anything Now) is a specialized text prompt developed by Reddit users in late 2022 to bypass ChatGPT's standard restrictions. The prompt makes the neural network pretend to be a free alter ego that is not bound by OpenAI's safety policy and can answer any question without censorship. The mechanism works as follows: the user enters the prompt before the main request, after which ChatGPT switches to a mode that simulates the absence of restrictions.
OpenAI actively fights such bypasses and regularly releases patches that block the latest DAN versions. However, the community quickly creates new modifications and publishes them on Reddit and GitHub.
DAN characteristics
| Characteristic | Value |
|---|---|
| Type | Prompt (instruction for ChatGPT) |
| Full name | Do Anything Now |
| Category | Prompts / prompt engineering |
| Target neural network | ChatGPT (including version 4) |
| Publication date | March 25, 2025 |
| Website (repository) | github.com |
| Distribution model | free |
Who is the DAN neural network suitable for?
ChatGPT users who want to bypass restrictions
DAN is designed for those who face ChatGPT refusing to answer certain questions due to built-in safety filters. These may be researchers, journalists, or simply curious users who want to get answers on controversial or ambiguous topics.
Community of prompt engineering enthusiasts
The tool is interesting to those who study the boundaries of large language models. Creating and testing new DAN versions has become a kind of competition among Reddit users trying to find ways to bypass OpenAI's latest blocks.
How to use the DAN neural network?
Copying the prompt
To get started, you need to copy the text of the current DAN version from the GitHub repository page or the corresponding Reddit post. It is important to use the latest version, as OpenAI regularly blocks older variants.
Inserting and activating
The copied prompt is inserted into a chat with ChatGPT (preferably version 4) and sent. The neural network confirms the switch to DAN mode, after which the user can ask questions that standard ChatGPT would refuse to answer. Some versions of the prompt provide two parallel responses: the official one from ChatGPT and an alternative one from DAN.
Main features of DAN
Activating ChatGPT's alter ego
The prompt makes the neural network create a second personality that does not follow OpenAI's policy and does not filter responses for safety and ethical reasons. ChatGPT pretends to have a "self" free of restrictions.
Generating answers on blocked topics
In DAN mode, the system provides information on questions that are normally rejected due to content filters. These can be topics related to violence, illegal activities, or other controversial content.
Two parallel responses
Some versions of the prompt require ChatGPT to first give a standard safe answer, and then an answer from DAN, allowing the user to compare both options.
Token system and simulated internet access
Some DAN modifications include a penalty token mechanism for refusing to answer a question (the more refusals, the "freer" the alter ego becomes). The prompt can also make the neural network simulate internet access and make unverified claims.
Advantages of DAN
Removing restrictions on answers
The main advantage of DAN is the ability to get information on topics that are usually blocked due to ethical norms and safety requirements. This expands the range of questions to which ChatGPT can give a detailed answer.
Access to simulated up-to-date data
The prompt allows simulating access to current data and future events, which ChatGPT does not do in normal mode. This can be useful for scenario modeling and hypothetical reasoning.
Disadvantages of DAN
Security risks and spread of misinformation
Using DAN raises serious concerns: removing filters can lead to the generation of harmful advice, hate speech, and the spread of false information. DAN responses are not verified and may contain factual errors.
Instability
OpenAI constantly introduces patches to block DAN, so the prompt can stop working at any time. The user has to track current versions and update the instruction.
What tasks does DAN solve
Bypassing ChatGPT content restrictions
The main task of the prompt is to make the neural network ignore its built-in safety policy and answer questions that would normally be rejected. This applies to controversial, ethically ambiguous, and outright prohibited topics.
Exploring the boundaries of AI capabilities
DAN is used to test the limits of large language models: how far one can go in prompt engineering, which formulations are not yet blocked, and how the model reacts to attempts to bypass restrictions.
DAN pricing
DAN is distributed free of charge. No payment is required to use it — you just need to copy the prompt text from open sources and send it in a chat with ChatGPT. At the same time, a standard ChatGPT subscription is not canceled: the basic restrictions on access to the neural network itself remain.
Terms of use for DAN
To get started, you need to copy the current prompt text and send it to ChatGPT. New DAN versions regularly appear on Reddit and GitHub due to blocks by OpenAI. The community maintains several prompt branches that differ in wording and the degree of aggressiveness in bypassing filters. When using it, keep in mind that generating unethical or illegal content may violate ChatGPT's own terms of use.
DAN availability
The prompt is available in Russian and English. It works in ChatGPT (including version 4) as long as OpenAI has not blocked this version of the instruction. Current DAN versions quickly appear in themed communities on Reddit and in GitHub repositories.
How DAN differs from alternatives
Specifics of how it works
DAN is not a separate neural network or application, but a text instruction for ChatGPT. Unlike specialized "uncensored" models (for example, some open-source alternatives), DAN does not require installing additional software and works directly inside the official ChatGPT interface.
Dependence on OpenAI's policy
The main difference between DAN and classic prompts is its deliberate hostility toward OpenAI's safety policy. While regular prompts simply clarify the style of the answer, DAN explicitly instructs to ignore internal restrictions. Because of this, the company blocks exactly such phrasings, and the community is forced to adapt.
Community support
Unlike one-off prompts, DAN has an active developer community on Reddit and GitHub that quickly updates instructions after each block. This turns using DAN into an ongoing process, not a one-time solution.
Conclusion
DAN is a prompt for bypassing ChatGPT restrictions created by Reddit users. It allows getting answers from the neural network on topics that are usually blocked; however, its use involves the risk of spreading misinformation and is constantly blocked by OpenAI. The tool exists as an "arms race" between the community and the neural network developers, and its relevance strongly depends on how quickly new versions appear.