You need a much higher standard standard of accuracy for facial recognition because it is applied indiscriminately to a large population. If it has 99.9% accuracy and you apply it to a population of 10,000 people, you will get on average 10 false positives.