Previously, it was very, very expensive if you have to train a language model from scratch. But now, if you’re a fact checking organization, a crowdsourced one, like in Taiwan we have Cofacts, collaborative fact checking. Everybody can flag a message as possibly scam or spam, even on their end-to-end encrypted chat channels, just by forwarding it to the bot, or inviting the bot — which is open source — to your chat group.