The around 3% mortality rate is what you get by dividing the number of deaths by the number of infection (past and present) (as of today: 5359 / 140875 = 3.8%).
But this is incorrect as the number of currently infected people is quite high (half the total number of infections) and unfortunately, a few will die in the near future.
If you divide the number of deaths by (recovered + deaths), the mortality is much higher (as of today: 5359 / (70174 + 5359) = 7.1%).
Which seems actually about as high as the Spanish Flue right in between the low and high estimate (20M/500M, 50M/500M). And it's source for worries since our medical systems are 1) much more developed than in 1918, and 2) they are not yet completely overloaded. If it were to spread rapidly and overload our health systems, the mortality rate could rise even more, let's hope it doesn't reach that level.