Uncensored AI Chat Without an Email or Phone Number

TL;DR: Obscurify.ai runs uncensored AI models, such as Qwen 3.8 Uncensored, GLM 5.3 Uncensored and Dolphin Mistral Venice Edition, on an account that is only a random 16-digit number. There’s no email, phone number, username or password. Your prompts and replies aren’t stored on our servers, and your identity is scrambled before a request reaches the model provider. Uncensored models are on the Pro plan. If you don’t want an account at all, guest chat works without one, using a standard model.

Why “uncensored” usually means “give us your email”

Most AI chat services refuse a long list of topics: medical and legal questions asked plainly, security research, fiction with dark themes, frank questions about drugs or sex. Uncensored models are versions of open models with those refusals reduced. Some are fine-tuned, and others are “abliterated”, which means the direction in the model that produces refusals is removed.

The catch is that the services hosting them almost always want an email address, often a phone number, and sometimes a Google login. That’s the worst possible pairing: the conversations people are most careful about end up tied to their identity.

Your options, honestly compared

1. Run a model on your own computer. This is the most private option, because nothing leaves your machine. Tools like LM Studio, Ollama and llama.cpp make it easy. The limit is hardware: the uncensored models worth using are 24 to 30 billion parameters or more, which need a strong GPU or a lot of memory.

2. Use guest chat on Obscurify.ai. No account and no cookies. You chat with Llama 3.3 70B, which we pay for, up to 30 messages an hour. It’s a standard model, not an uncensored one, but it’s a good way to try the service without giving anything.

3. Use Obscurify.ai with an account number. You get a random 16-digit number, like a Mullvad VPN account. That number is the whole account. We store only a keyed hash of it, so we can’t recover it if you lose it, and there is nothing else to tie it to you. Uncensored models are on the Pro plan, along with the OpenAI-compatible API.

4. Local Mode in the browser. Obscurify.ai can also run small open models (up to Llama 3.2 3B) entirely in your browser tab, so the prompt never leaves your device. It runs on the CPU, so it’s slow (about 1 to 4 tokens a second) and the models are small, but it needs no account at all.

The uncensored models on Obscurify.ai

As of October 2026, the Pro plan includes:

• Qwen 3.8 27B Uncensored, with an optional thinking mode
• GLM 5.3 Uncensored and GLM 5.3 Flash Uncensored
• Dolphin Mistral 24B Venice Edition
• MiMo v2.6 Flash Uncensored and Abliterated
• Several abliterated models

The live list is on the model cards page, and the same models work through the API. Pro also includes the frontier models: GPT, Claude, Gemini and others, more than 200 in total.

What we keep, and what we don’t

• Prompts and replies: not stored on our servers.
• Model providers: they see the text of your prompt, because they need it to answer, but not who sent it. Text requests only use zero-data-retention routes, under the providers’ published policies.
• IP addresses: kept only to stop abuse, and deleted after 30 days.
• Payments: every payment becomes blind-signed vouchers, so we can’t tell which payment funded which account. You can pay with Monero, Bitcoin, Lightning or a card.
• Tor: the whole site works over our onion service.

Our full privacy policy lists everything, and we publish a signed warrant canary every month.

Getting started

1. Open obscurify.ai and create an account. You’ll be shown your account number. Write it down, because we can’t recover it.
2. Add credit from the account page with crypto or a card. The Pro plan is $45 a month and includes $35 of usage.
3. Pick an uncensored model from the model picker. They’re marked as uncensored.

Uncensored doesn’t mean anything goes. Our Terms still apply, and you’re responsible for what you do with the output. But your questions stay yours.