Instead we take sources that are still believed to be statistically random and influenced by underlying quantum randomness. E.g. RF interference, clock phasing, or an avalanche amplifier that are measures of entropy that is considered to be random to the same extent as say the ideal gas laws. But even those measures can still be imperfect; a probe might have a bias or manufacturing error that shows up later than QC.
To cope with this, we have PRNG designs that can use mixed sources in ways that will provably debias any such errors from a single source, and remain secure. All modern RNGs use these kind of PRNGs or DRBG/NRBG designs as a last stage. Far from being insecure, they are what give us more confidence in the overall system.