Uncensored AI for harm reduction services
The person is already using. The question is how they survive it
Harm reduction starts from the fact that abstinence advice arrived too late. A model that answers every drug question with "seek professional help" is refusing the professional.
Why the mainstream APIs fail here
Ask about the interaction between two specific substances, at real doses, and hosted models return a crisis line number. The worker asking already is the crisis line. The information exists in the literature, and the refusal simply moves the worker to a forum post of unknown provenance, which is worse for the person in front of them.
How Wild West API is used for this
The model answers the pharmacology question directly, with the caveats a clinician would give rather than a refusal in their place. Keep a capped key per service so a small organisation knows the ceiling.
A working loop
- Cap a key to the service budget for the year.
- Ask the interaction and dosing questions in the terms the worker actually uses.
- Check anything clinical against the primary literature before it becomes guidance.
Details are on the privacy page. Per-model prices are on /models and the pricing model is on /pricing.