I've worked with a lot of very good facial recognition technology. They all have demonstrated biases for white males. Even in the ideal scenario they are far too inaccurate (<97%) to be used for identification, unless your false reject rate is extremely high (>30%). And if it is a less ideal scenario (elderly Latina, Asian woman, etc) you'll be lucky to get 80% accuracy without high false reject rates.

When it comes down to it, accurate facial recognition is at best 99% accurate. But being wrong 1 in 100 times is a horrible number for an authentication system. Anything else you ever use for authentication (password, iOS face scanning, PIN, etc) would be considered a complete joke if it only got to 99% accuracy.