They are claiming that this is "snake oil technology". I think that research is exactly what's needed to determine whether using AI for facial lie detection can work or not and, if it does, how well.

I don't think this is a waste of public money to research this and if it works it could be very useful indeed in cases where asserting the truth is important. They mention immigration checks as an example but police interviews are another obvious application among many.

Of course, like for new technology we would need to think about the need for regulation and appropriate procedures.