EU customs authorities are a cornerstone in the international trade landscape, playing a critical role in preventing the entry of illegal goods, safeguarding revenue, & ensuring the seamless flow of goods. Balancing these responsibilities is particularly challenging in today’s environment, where the volume of global trade continues to expand rapidly. Mitigating the entrance of illegal goods is becoming increasingly difficult, especially with limited human resources. Aiming to address these challenges, the CustomAI consortium has united its expertise and competences to develop an AI-toolkit that will reduce the number of false positives (situations where the cargoes like shipping containers or parcels have been selected for inspection despite not containing contraband). The proposed AI-toolkit will revolutionise customs operations by involving non-intrusive and robust AI-enhanced technologies for predicting, detecting, and selecting high-risk cargoes for inspection. The VCCO concept is adopted for managing all processes in the customs control of artefacts (e.g. container, parcel). Key components of the AI toolkit include: * AI-based risk anticipation relying on AI-analysis of internal knowledge in compilation with external multilingual data, including manifest and declarations. Only relevant cargos will be sent for inspection. * AI-enhanced vapour-based detectors implied only on the containers selected in the previous step. * AI-based x-ray for threat detection in containers applied on output of step two (the human inspection takes place only after this step). * Multimodal LLM Continual Learning model, which will have as input, x-ray and camera images, and will be trained on threat dataset composed of threat samples (x-ray and visual images of threat parcels) updated by customs. * Blockchain technology for secure data sharing & supply chain traceability. By adopting these cutting-edge technologies, the CustomAI toolkit is set to revolutionise customs operations.

By combining IoT solutions with edge-based AI, our project PACK significantly decreases the number of trucks to be stopped for checking, enables fully automated customs checks, improves energy efficiency by optimising the equipment to operate autonomously with minimal environmental impact. It aims to deliver a highly secure, cyber-resilient access control system specifically designed for the cargo compartments of freight trucks. This innovative system enhances and integrates advanced detection, tracking, and risk-based selective control systems, featuring trustworthy algorithms that reduce false positives and biases. PACK brings more efficient customs control processes by a smart IoT-enabled customs seal to deter tampering and an intelligent IoT sensor array equipped with edge-based AI for continuous monitoring. With this setup, the system can independently evaluate the need to issue alerts to supervising authorities, thereby streamlining roadside inspections. Additionally, it allows customs and law enforcement agencies to access and unlock the customs seal at any point during transit, including locations outside traditional customs yards and checkpoints. Our goal is to demonstrate that implementing PACK across the European Union can significantly reduce freight truck waiting times at external EU borders, by just a 10% market penetration saving the logistics and transportation industry ~ €241 million and 2.35 million hours of time annually. This improvement would also result in the conservation of ~70.5 million litres of fuel annually, reducing CO2 emissions by approximately 190.000 metric tons each year. The complete solution will be validated with end-users in realistic operational scenarios, running parallel to actual customs inspections to benchmark its performance and deliver measurable results. Market uptake is planned in a public-private cost sharing model where customs authorities, truck operators and logistics companies share deployment costs.