I think it’s so strange how Pro wasn’t launched for Bard in Europe yet. I thought Bard was already cleared for EU use following their lengthy delay, and that this clearance wouldn’t be a recurring issue to overcome for each new underlying language model. Unless it’s technically hard to NOT train it on your data or whatever. Weird.
I suspect this is because inference is very expensive (much like GPT-4) and their expected ARPU (average revenue per user) in Europe is just not high enough to be worth the cost.
Bard is explicit "humans will have access to your data and use it to improve our service", without even the opt-out that ChatGPT has, so, yeah, given their desire to catch up, not spending money to deploy it places where there is any kind of meaningful protection of personal data that might conflict with that makes sense.
