I think there’s a codec squishy enough that it can squish voice into, say, 9600bps, and I think there’s a smartphone fast enough to intercept telephone audio, decode it in regular software at 9600bps, and stuff it back through that codec.
Of course, since it’s just bits, this allows strong encryption between both parties. It would also be possible to build boxes – essentially 9600bps modems with goodies – capable of allowing landline callers to interact securely, over the old-school copper phone network, with users of encrypted calling apps on smartphones. This would close a surveillance hole – the PSTN – where it’s still casually easy to eavesdrop.
I think the boxes for landlines, if supported by 2 major apps, could cause an ad hoc standard to emerge, which would be good.