A six-month pilot program will see an automated border agent question travelers on non-EU borders in Hungary, Latvia, and Greece, Gizmodo reports.
The system will ask travelers security questions, such as questions about what’s in their luggage, while scanning their faces for “micro-gestures” that can indicate if they’re lying. The bot will be calibrated based on people’s ethnicity, language, and gender.
It’s part of an EU-funded experimental system called iBorderCtrl. A test of an earlier system by some of the same researchers saw a roughly 76% accuracy rate at detecting people who lied or told the truth at experimenter request, though the team thinks they can get the accuracy up to closer to 85%.
To start out, the system won’t unilaterally block anyone from entering the EU, though it may send some people off to human agents for further assessment. The test comes as countries around the world are increasingly deploying biometrics at airports and border crossings, raising concerns about accuracy and privacy.