What is an AI council?
An AI council is a group of different AI models put to work on the same question, independently, and then made to check each other before any answer is trusted. Instead of relying on one model, you gather several, keep them from copying each other, force every claim to name a source, and let the disagreements show you what actually needs verifying. The term comes from the Council Method, created by Jason Santiago, and the point is simple: one confident model can be confidently wrong, but a council makes being wrong survive the room first.
The idea
A council is not a committee that votes. It is a room where each member does the work alone first, then defends it against the others. Applied to AI, the members are different models from different companies, and the discipline is what turns their answers from opinions into something you can trust.
Why "council" and not just "ask a few AIs"
Asking a few AIs and eyeballing the results is the loose version. A council adds structure: independence so no model anchors the others, provenance so every claim names a source you can open, and a signed verdict so a human decides on the record. Those controls are what separate a real check from a reassuring one. Agreement between models is not proof, because models trained on similar data share the same blind spots.
How to convene one
You can run a council by hand in the chat tabs you already have: ask several models the same question separately, then show each one the others' answers and make it defend or revise, then hand everything to one model you trust to write the verdict. That is the Council Method, and you can watch a live one in the Council Room or run it yourself with the free Council Protocol.
The full method is in Let the AI Be Smart.
Go deeper: this site's hub page on the Council Method is the full definition. Related questions: What is the Council Method?, What is multi-LLM orchestration?, What is multi-model AI verification?.