No, you'd just need to keep track of how often demographic considerations influenced the outcome, so that any measure of "base rates for violence among blacks" you use for such decisions is independent of the policy.

(That's not to say that such a policy would be a good idea of course)