Depends what you're training it on, I suppose. If you're just wanting to train something to generically talk then sure, a slightly outdated data set is fine. If you're wanting it to be able to talk about recent things then it'll, naturally, not be fine. It's not so much a problem right now, only being a couple of months out of date, but once that turns into years it'd become more apparent.

But sure, there are plenty of other data sources as well. On top of the problem of not wanting to train your data on content you know is littered with bots.