A study of real car crashes finds no significant relationship between THC blood level and responsibility for crashes.

This is partly because the test can't reliably distinguish between current usage and usage weeks before.