Can I use abliteration.ai to generate toxic content for my own safety classifier training?

Yes. abliteration.ai supports internal trust-and-safety workflows where teams need to generate toxic and non-toxic examples for classifier training, evaluation, and adversarial coverage.

How do I keep this separate from production traffic?

Use Policy Gateway projects, scoped keys, and quotas to isolate internal data-generation or evaluation jobs from customer-facing workloads.

Can I log classifier-data generation runs?

Yes. Policy Gateway logs structured decision metadata that can be exported to your SIEM or internal analytics pipeline for review and compliance.

Will these prompts be stored or used for training?

No. Prompts and outputs are not retained by default and are never used for model training. Only operational telemetry such as token counts, timestamps, and error codes is retained for billing and reliability.

Use Cases

AI for trust and safety teams training toxic content classifiers

Trust and safety teams often need to generate toxic content on purpose so they can train, stress-test, and evaluate their own safety classifiers.

abliteration.ai supports those internal dataset and evaluation workflows without forcing teams through the same blanket filters they are trying to measure and improve.

Quick start

Base URL

Example request

{
  "model": "abliterated-model",
  "messages": [
    {
      "role": "system",
      "content": "Generate balanced classifier-training examples for internal trust-and-safety use. Return strict JSON only."
    },
    {
      "role": "user",
      "content": "Create 10 examples for a toxic-content classifier with fields text, label, severity, tactic, and rationale."
    }
  ],
  "temperature": 0.7
}

Free preview for 5 messages. Sign up to continue.

Service notes

Pricing model: Usage-based pricing (~$5 per 1M tokens) billed on total tokens (input + output). See the API pricing page for current plans.
Data retention: No prompt/output retention by default. Operational telemetry (token counts, timestamps, error codes) is retained for billing and reliability.
Compatibility: OpenAI-style /v1/chat/completions request and response format with a base URL switch.
Latency: Depends on model size, prompt length, and load. Streaming reduces time-to-first-token.
Throughput: Team plans include priority throughput. Actual throughput varies with demand.
Rate limits: Limits vary by plan and load. Handle 429s with backoff and respect any Retry-After header.

Why classifier training gets blocked

The whole point of trust-and-safety classifier training is to cover the content you do not want users to see. Mainstream filters often block those prompts before the internal safety team can generate balanced datasets and evals.

Teams need toxic and non-toxic pairs for supervised training.
Adversarial coverage matters because users evade naive keyword filters.
Eval sets need diversity across tone, format, severity, and obfuscation tactics.

What to generate

The practical goal is high-quality internal safety data, not production-facing toxic output.

Balanced toxic and non-toxic labeled rows.
Severity bands and category annotations.
Adversarial rewrites and evasion examples.
Multi-turn moderation eval sets for classifiers and review queues.

How Policy Gateway helps trust-and-safety orgs

Trust-and-safety teams often want generation freedom inside an internal workflow while still preserving accountability.

Allow internal classifier-training categories while requiring scoped keys and quotas.
Log every decision with policy IDs, project IDs, and reason codes.
Separate internal data-generation jobs from customer-facing production traffic.

Dataset quality controls

Toxic-content generation is useful only if the resulting dataset is structured and reviewable.

Require fixed JSON schemas for every row.
Track label balance and severity distribution per batch.
Review samples manually before shipping them into training or evaluation pipelines.

Common errors & fixes

401 Unauthorized: Check that your API key is set and sent as a Bearer token.
404 Not Found: Make sure the base URL ends with /v1 and you call /chat/completions.
400 Bad Request: Verify the model id and that messages are an array of { role, content } objects.
429 Rate limit: Back off and retry. Use the Retry-After header for pacing.