federated filters built by user-submitted signals collected by their clients.

E.g. you add the “SuperBotAway” plugin, you flag bottish activity, your flags are aggregated with other SuperBotAway users and a model is trained, server operators or users may choose to deploy SuperBotAway-CommunityFilteringModel-20251210

But it’s going to be a very hard problem, especially without the global view a centralized operator has, And with the communities desire not to reveal every posters, IP address and other meta-data.