In the study properties, random vibration options, what upper limit are you using? You probably want to set that to 2000.
You could also define your excitation curve for points beyond 2 kHz. Either of these two solutions would solve your error message.
I do see, though, that your first vibration mode is listed at 10 kHz. That's way above your input curve, so you won't really be getting any useful data out of this study. Is your model really that stiff, or might it be over constrained?