First of all, one should differentiate between force fields, i.e. parameter sets, and program versions.
As Stéphane pointed out, Amber12 contains a new parameter set parm12, which is based on the parm99SB parameter set, which in turn is an extension of the parm99 parameter set. However, I did not find any literature reference in the Amber manual some time ago, where this new parameter set is described (maybe this has changed in the meantime).
I can only speculate, why the development of new parameter sets within the inner Amber developers community rely on the parm99SB parameter set. On of the main reasons certainly is the huge amount of studies published with parm99SB, so the successful application of that parameter set is demonstrated at length.
The parm03 set on the other hand, which contains charges parameterized using implicit solvent calculations (see Stéphane's answer), is rarely used, like is the parm02 set, which contains polarizabilities and thus fluctuating charges.
The developers of parm12 obviously wanted to inherit from the successful parm99SB parameter set.