This suggests that we are pretty close to not being able to rely on voice biometrics.

Does anyone in the field know more?