Abliteration AI

Abliteration AI is a Palo Alto company selling an OpenAI-compatible LLM API for work most assistants turn down: AI red teaming, cybersecurity, trust and safety, synthetic data and evals, ML research, and defense or government workflows. Its models start from open-weight releases that are abliterated, a weight-editing technique that removes the internal direction behind refusals, and then fine-tuned for that work.

The pitch is unrestricted but not ungoverned. Customers set their own limits through a Policy Gateway with policy-as-code and per-project keys, and the company says it retains zero data: prompts and outputs are never stored or logged.

On Venice the lab has a single model, Abliterated Large V2, a text-only reasoning model built on GLM-5.3 with a one-million-token context window. It follows Abliterated Large, the GLM-5.2 version the company launched in July 2026.