More than 700 million people enter the EU every year – a number that is growing rapidly. The huge volume of passengers and vehicles exert pressure on external borders, making it increasingly difficult for border services to maintain strict security protocols – to check the travel documents and biometrics of each passenger while minimising infringements.

Initial tests for the implementation of an artificial intelligence system at EU border crossings will start trial very soon. The system is an advanced lie detector and will be used for six months in four points in Hungary, Latvia and Greece, where there is a border with countries outside the EU.

Called iBorderCtrl, it is a project financed by Brussels, which uses AI to ease the border process and make passage faster.

Passengers must fill in an online form and upload some documents as their passports, then a virtual border police officer begins to ask them questions. Among the most common enquiries are “what you have in your briefcase?” and “If we open them, will the content be the same as what you tell us?” On these issues, travellers will respond to a webcam and the system will analyse and evaluate tens of micro-gestures.

If iBorderCtrl determines that the person tells the truth, the system issues a QR code with which he/she crosses the border. If there is suspicion that he/she lies, its biometric information is collected – including footprints, facial expressions and data from the palm of the hand – and a real policeman is called to continue with the interrogation and take a final judgment.

The programme is still considered as a strictly experimental and in its current form cannot prevent the crossing of the border. Early tests at its previous version ended with 76 percent success – promising but insufficient. A team member behind iBorderCtrl says that they are “quite confident” that the success rate can increase to 85 percent.

But even if this goal is achieved, it leaves enough room for mistakes. Overall, however, this is a problem for most facial recognition algorithms, which still show significant deviations and even bias. Nevertheless, the European Union has committed hopes in this direction and even finances the project under its Horizon 2020 programme for innovation.