Administrators choose the model for each task. Some moderation paths can automatically flag content, restrict accounts and conditionally unpublish posts. The source proves those capabilities exist, not that any particular live Forem site has enabled Jev.
How you can use it
Have a developer try one lower-risk task, such as article tags, in a separate Forem installation before changing a community's moderation. An administrator selects Jev for each function and supplies a TypeSafe access key. Review the suggested labels against actual examples; the model's answer does not prove the label is right.
Forem sends content and sometimes profile details to TypeSafe, and retains request and response bodies in its audit records. Its extra spam-screening step only requests another check, but other paths can block accounts automatically. Review privacy, mistakes and recovery before enabling those paths; safety was not demonstrated by this source review.