No. It produces ranked candidate matches that must be reviewed and independently corroborated before anyone is treated as identified.
NIST testing found accuracy varies between algorithms, with some showing higher error rates for certain demographic groups.
Greatly. Blurred, angled, poorly lit or low-resolution old photographs produce much less reliable results than modern frontal images.
Yes. Several U.S. wrongful arrests have been linked to over-reliance on facial recognition without adequate further investigation.
No. Such tools can misidentify people; pass potential leads to investigators instead of publishing them.