Full Disclosure:
I was a very active member of the NTP Hackers team for 25+ years, network time protocol did foresee the possibility of both positive and negative leap seconds. Among other things, there's a two-bit field in every server packet where two of the four combinations indicate an upcoming plus or minus leap second, to be applied at the next UTC midnight.
That said, I am 99%+ sure that we will _not_ subtract any leap seconds in the next 100 years, and the reason is very simple:
Too many systems would fail very badly!
Over longer time scales, the Earth's rotation _will_ slow down further, adding positive leap seconds, and correcting this more short-scale speedup would therefore happen automatically. However, this is almost certainly moot because we will abolish leap seconds completely, and instead require astronomers and others who care a lot about the offset between UTC and UT1 to maintain their own tables.
Currently, GPS sats all transmit the UTC-UT1 offset in a field which can only cover about a second, so what we have been doing over the last 40 years is to measure said offset and when it got above ~0.6 seconds, the IERTS would announce a positive leap second so that the offset became -0.4 instead.
Terje