You need to know the false negative rate to evaluate it against a coin flip. For example, a hypothetical machine with a 50% false positive rate but 0% false negative rate could still be really useful, if the machine itself is cheap/fast and the subsequent screening slower/more expensive. With such a machine, you know that the 50% who pass are clean, thus reducing your followup screening load by half.

Of course, a machine with ~50% false positive and negative rates is indistinguishable from flipping a coin.