In what ways could Artificial Intelligence make Customs NL smarter?

Customs NL is always looking for new ways of conducting its supervision and reducing the inspection burden for trade and industry. For this purpose Customs tests groundbreaking technologies, working methods and partnerships. By way of example, studies are currently conducted to find out to what extent Artificial Intelligence (AI) can make the organisation smarter.

Computers have a computing capacity that far exceeds the mental powers of humans. Within seconds, they can identify connections and patterns in a heap of information which would have remained hidden to the human eye otherwise. Hence Customs more and more use computer technology to find the answers to complex issues. Specially-developed, mathematical models are given a highly specific instruction and a set of data from which a solution must be deduced. This method practically always produces a useful outcome, but the disadvantage is that they are time-consuming in terms of preparation. It would be much more efficient if an electronic system, of its own accord, initiates a specific search in a wide range of information sources, filters out what it believes to be the main knowledge on a certain subject and subsequently presents this knowledge. This is also referred to as artificial intelligence, i.e. machines that think and teach themselves. Customs are particularly interested in this phenomenon and are exploring the possibilities thereof within the market sector and the academic world. It is mainly deemed as a potentially useful tool in making the right choices at an operational level.

In modern-day practice, Customs are confronted with an endless variety of goods. Some of those are banned, e.g. because they form a threat to public health or public safety. Others are obliged to have a permit, such as products that are subject to sanctions legislation, whereas another group may be subject to import duties, consumption tax, excise duty or other levies. Despite their training and many years of experience, customs officers cannot always easily determine the status of each consignment they come across. Suppose a control officer comes across a batch of jeans with labels showing Chinese characters. In that case, he will check, among other things, whether the price stated on the cargo documents is realistic. That seems to be a straight-forward question, yet the background is extremely complex. This may involve low-value textiles, for example, which are subject to an anti-dumping levy once they enter the EU. However, this product group has numerous sub-categories, as a result of which classification in the correct rate band is often not a sinecure. Therefore, the officer in this example would benefit from an IT application that quickly provides him with clear information on the correct classification of the jeans concerned, preferably also on the basis of recent European case law.

In simple terms, what Customs need is a central, electronic brain. An artificial neural network with access to all the knowledge ever gained and digitalised by the agency and to a wide range of open sources at the same time. One that does not inundate the user with a wave of information, but one that offers bite-size chunks of useful data surrounding a predefined theme. Not only will such a system be able to read texts at high speed and recognise relevant terms therein, but it will also be able to interpret the context and draw conclusions from it. In order to ensure its proper performance, sophisticated algorithms need to be developed first and clear definitions of concepts drawn up – a time-consuming process. The application of Artificial Intelligence is, partly for that reason, still a distant ambition.

Nonetheless, the agency recently recruited freshly graduated specialists who, within test environments, test what AI can bring to the organisation – in collaboration with private parties and research institutes. This involves the assessment of a wide variety of options, such as automated image recognition with camera surveillance of travellers at Schiphol Airport. Staffing of the control room at the airport is not such that all recordings can be monitored in real time and at a continuous level of accuracy. Hence it would be useful if a smart system looks over the shoulder of on-duty personnel, alerting them to any suspicious activities, such as the transfer of an item of luggage between individuals within the flow of passengers. Another option concerns support in interviews with travellers, e.g. during 100% inspections of high-risk flights. Special interrogation software could flag up any inconsistencies in the answers of passengers to customs officers who can direct interviews in that manner. It could considerably increase the chances of catching drug smugglers for instance.

It is clear that, in the longer term, AI could make customs surveillance even smarter, stricter and quicker than it is today. This helps the agency to realise its strategic objectives: promoting correct and full payment of taxes, protecting society and strengthening the economic competitiveness of the Netherlands and Europe. Still, it needs to be said that no miracles can be expected from technology alone. Despite the growing importance of automation, humans will continue to play a role of importance. Customs offers with the necessary expertise will always assess the suggestions put forward by the systems.

Share this post