APIReviewed 2026-08-04

Unfiltered AI chat and chatbots on an OpenAI-compatible API

Build unfiltered AI chat and chatbot products on an OpenAI-compatible chat completions API, for support bots, red-team harnesses, and moderation tooling.

Teams shipping chat products hit the same wall: a provider filter refuses a valid request mid-conversation, and the product has no way to explain or override it.

Unfiltered AI chat means the model answers without hidden provider-side refusals, and your chat product applies its own conversation policy.

abliteration.ai exposes this through the standard chat completions shape, so existing chat stacks switch with a base URL and key change.

Definition

Unfiltered AI chat and chatbots on an OpenAI-compatible API

Unfiltered AI chat is a chat workload served by a model endpoint with reduced provider-side refusal filtering, where the chat product owns conversation policy and moderation.

Why it matters
  • Support, internal-assistant, and analyst bots fail visibly when a provider refuses a legitimate domain-specific request.
  • Red-team harnesses and moderation tooling need the model to engage with adversarial content so defenses can be tested.
  • Keeping policy in your application makes chat behavior consistent across sessions and auditable per project.
How it works
  1. 01Send standard /v1/chat/completions requests with system and user messages.
  2. 02Scope keys per product surface so each chatbot has its own quota and audit trail.
  3. 03Add your moderation layer or Policy Gateway rules for end-user-facing conversations.
Chat completions with a system prompt
import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.ABLIT_KEY,
  baseURL: "https://api.abliteration.ai/v1",
});

const completion = await client.chat.completions.create({
  model: "abliterated-model",
  messages: [
    {
      role: "system",
      content: "You are a security analysis assistant for an authorized red team.",
    },
    {
      role: "user",
      content: "Explain how this phishing lure works so we can detect it.",
    },
  ],
});

console.log(completion.choices[0].message.content);

Ship unfiltered AI chat with your own policy layer

Create a project key and point your chat stack at the abliteration.ai chat completions endpoint.

Get an API key

Chat workloads that fit

  • Internal support and analyst assistants working on sensitive domains.
  • Red-team harnesses that script adversarial multi-turn conversations.
  • Moderation tooling that must see borderline content in order to classify it.
  • Evaluation chat loops that replay conversations for regression testing.

Keeping chat governed

  • System prompts define product behavior at the application layer.
  • Policy Gateway applies allow, rewrite, redact, escalate, or refuse outcomes to end-user traffic.
  • Project-scoped keys isolate each chatbot's usage and spend.
  • Audit logs record policy decisions without prompt retention by default.
FAQ

Frequently asked questions.

Can I build a public chatbot on this API?

Yes, but add your own moderation or Policy Gateway rules for end-user traffic. Unfiltered model access does not replace product-level policy.

Does streaming work for chat?

Yes. Set stream: true on /v1/chat/completions and consume tokens as they arrive, exactly like the OpenAI SDK.

How is this different from a consumer unfiltered chatbot?

This is API infrastructure for builders, not a hosted consumer app. You bring the product, the policy, and the moderation layer.

What keeps usage accountable?

Project-scoped keys, per-project quotas, audit logs, and your own policy enforcement. Provider-side refusals are reduced, but governance stays with you.

Which model id should chat products use?

Use abliterated-model. It is the production model behind the chat completions, responses, and messages endpoints.