Frontex, the European border and coast guard agency, is responsible for the control and management of the external borders of the Schengen area. As part of its digital transition strategy, it has published a report on a study carried out around the main requirements and opportunities existing around the adoption of AI systems to better manage borders. This research was made possible by data provided by RAND Corporation.

The objectives of the study conducted by Frontex

The study seeks to explore the benefits and potential areas of adoption of AI. It aims to gain a better understanding of its capabilities in order to exploit it in the daily missions and tasks carried out by coast guards and border guards. Frontex is aware of the technological barriers that may exist and hopes, with the help of this report, to be able to overcome them.

This study has sought to answer four fundamental questions:

• What is the current landscape of AI application to border security?
• What innovative AI-based systems could be applied to border security?
• In which areas of border security could these AI-based systems be applied?
• What steps are needed to integrate AI-based systems into border security?

Potential uses of AI for Frontex

The potential uses of AI put forward by Frontex group several border security functions:

• Situational awareness and analysis: AI systems can be used to collect, fuse and analyse data in real time or already stored to facilitate decision making and response performance in complex environments. Examples include monitoring people within a defined area, but also vehicles and objects. Systems mentioned that have the capability to perform such tasks are AI-enabled surveillance installations (surveillance towers) and autonomous systems (e.g. drones and networked heterogeneous robotic systems).
• Information management: Mainly data and information management through data mining and fusion techniques, NLP, image, text or pattern recognition, etc. The objective is to automate information (thanks to machine learning models in particular). Some processes, such as recruitment, can use AI to manage data.
• Communication: AI offers communication and information sharing capabilities, including in authentication technologies. There are other capabilities driven by NLP techniques (e.g. chatbots).
• Detection, identification and authentication: AI could be used to detect and identify threats and authenticate people and objects. This includes automated border control using AI, biometric scanning, facial recognition, and the creation and analysis of authentication documents (passports, visas), as well as threat detection capabilities using object recognition and cognitive robotics (e.g., creating robotic border patrol agents).
• Training and exercise: AI can support training and exercise through simulated real-world situations (creating simulated environments through virtual or augmented reality).

Adoption of AI-based systems: potential benefits and barriers

According to the study, several technological and non-technological factors could act as potential barriers to using AI systems:

• Technological barriers: AI systems can be operational and function if they are trained with sufficient data. If there is not enough data, the development of such models would be useless and the model would be ineffective.
• Commercial barriers: The use of AI models has a significant cost, although the price has been falling in recent years.
• Understanding and knowledge of AI: The use of AI in border security would be a new application in Europe. Some of the tasks that could potentially be performed by an AI system may not have been previously conceived and will need to be developed specifically for Frontex.
• Ethics: The algorithms used must respect human rights and may, at times, entail risks that impede these rights. These models must also respect the terms of the RGPD.

or alternatively as enablers:

• Technological leverage and iterative development: The adoption of AI systems could lead to a significant advance in this area. As explained above, the application of AI in the field of border security is not well exploited in Europe. Some systems could benefit from this lever: neural networks, sensory computing, blockchain integration, cognitive robotics, etc.
• Improved user platforms: Interfaces could also be more intuitive (in biometric scanning or surveillance technologies), despite the fact that AI systems are becoming increasingly complex.
• Democratization of AI: The commercialization and democratization of AI models will potentially contribute to lower production costs, improving the economic viability of the sector.
• Public awareness: Increasing use of AI to provide a variety of services can enhance the positive image of AI to the general public.

Let’s note one last point, not mentioned in the study: that of the brand new proposal for European regulations on artificial intelligence that Frontex will have to take into account.

Translated from Intelligence artificielle pour le contrôle et la gestion des frontières : Focus sur le rapport de Frontex