Q {0.18}

Abliteration joined a16z @speedrun. You control the guardrails. Unrestricted, not ungoverned.

Abliteration is an OpenAI-compatible API for less-restricted AI models with a built-in governance layer. Teams use it when mainstream models refuse too much, but raw access without controls is not an option. The policy layer lets you inspect and act on traffic around the model, including rewriting, redacting, escalating, or blocking outputs. Prompts and outputs are not retained by default, and decision metadata can plug into existing audit workflows. Best fit: security, trust and safety, synthetic data, and other high-control AI use cases. Unrestricted, not ungoverned.