Actually, “related” is correct, but I should have made it clearer. The intention here is to make your bot gracefully handle topics that are relevant but that your bot hasn’t been trained on yet. So, for example, a banking bot that is only trained to deal with credit cards will quite likely get asked about savings accounts or mortgages, because the user is talking to the bank’s bot. The bot should recognise these relevant topics and respond, say, with a link to the relevant part of their website.
Now, the bot should also handle unrelated topics, but this is a more general catch-all. Getting back “Sorry I don’t understand” from a banking bot when the user asked “Can I order a pizza?” is much more forgivable than, the same response to “What are your mortgage rates?”