Fool the Chat · Research consent

We may use votes and eligible items to train models that tell generated media from real captures. That is research, not the public score. Player screens do not say we are collecting training data. This page does.

The training checkbox on submit defaults to off. Account-level research consent also defaults off. The submit checkbox currently defaults to off.

Withdraw consent on your Profile: turn the research toggle off. Turning it off stops future grants. It does not rewrite Gold or Verified grants already recorded. Open Profile.

URL-discussion items (rights_status url_discussion_only) never become training labels.

Unresolved Help is never train_eligible. Gold items and Verified Challenges with a rights grant can be marked train-eligible.