8 - Big Data, AI and Border Enforcement of Intellectual Property Rights

Summary

Customs surveillance of intellectual property is an efficient way to quickly and effectively provide legal protection to the right-holder, as it allows to nip the infringements in the bud. Technology has drastically changed the means and mechanisms of customs enforcement, as it increases the possibilities of identifying and detaining goods infringing IPRs, and makes it more feasible to assess in advance where control is required.

However, assessing in advance and acting when appropriate does not always match well with fundamental intellectual property principles (territoriality), global trade norms (freedom of transit), global intellectual property rules, and due process requirements.

This chapter explores some of the challenges and opportunities brought by AI, big data and distributed ledger technologies to customs enforcement of IPRs. It looks at AI’s transformative influence on IP enforcement and the digitization and use of big data in customs control.

A Digitalization of Intellectual Property Enforcement

Customs surveillance of intellectual property (IP) is ‘an efficient way to quickly and effectively provide legal protection to the right-holder’,¹ since it makes it possible to ‘nip the infringements in the bud’.² The World Trade Organization’s (WTO) Agreement on Trade-Related Aspects of Intellectual Property Rights (TRIPS) introduced, back in 1995, the first comprehensive multilateral regulation on border measures applicable to goods protected by intellectual property rights (IPRs).³ Since then, global trade flows have increased,⁴ new trade agreements regulating IP have substantially modified international intellectual property law⁵ and, most importantly for this chapter’s discussion, technology has drastically changed the means and mechanisms of customs enforcement.

Technological developments increase the possibilities of identifying and detaining goods infringing IPRs, and make it more feasible to ‘assess in advance and control where required’, which is the ideal pattern of action from a customs’ risk management perspective.⁶ However, assessing in advance and acting when appropriate does not always match well with fundamental intellectual property principles (territoriality), global trade norms (freedom of transit), global intellectual property rules (Articles 51 and 52 TRIPS), and due process requirements. This chapter, in light of these basic norms and principles, explores some of the challenges and opportunities brought by artificial intelligence (AI), big data and distributed ledger technologies to customs enforcement of IPRs.⁷

B How Artificial Intelligence Transforms Intellectual Property Enforcement

Fuelled by a profusion of digitized data and rapidly advancing computational processing power,⁸ AI techniques and functional applications⁹ give rise to unprecedented opportunities of innovation and creativity, and also bring about important challenges to intellectual property protection.¹⁰ Three major areas stand out.

First, the use of AI techniques to generate innovative and creative products has prompted discussion concerning the need, extent and legal grounds for the IP protection of products developed thanks to AI.¹¹ Second, regarding the means that enable creation and innovation, the intellectual property protection of AI techniques and functional applications per se has skyrocketed in recent years.¹² Third, the use of AI, big data and distributed ledger technologies also impacts IP enforcement. The increase of available information and the changes in the techniques used to make such information useful impact fundamental aspects of intellectual property enforcement. These changes revolve around three concepts: authority, automation and centralization.

A substantial change concerns the authority in charge of law enforcement and the process of privatization of law enforcement by means of delegation of public authority.¹³ A telling example is that of online intermediaries,¹⁴ which have acquired a central role in managing behaviour in the digital environment.¹⁵ From free speech to access to information, many issues are mediated by search engines, websites and social networks. Their power also expands to intellectual property enforcement, since online intermediaries not only identify infringement, but also produce the information regarding the infringing activity, the infringers, the channels of commerce and its financial aspects.¹⁶ Intermediaries furthermore use algorithms to remove allegedly infringing content upon notice of infringement, as well as to ex ante monitor, filter, block and disable access to content automatically flagged as infringing.¹⁷

Privatization leads to the convergence of law enforcement and adjudication powers.¹⁸ Private stakeholders that identify infringing products also apply the corresponding sanctions of destroying infringing goods and cutting access to the Internet. The conventional way of functioning has been drastically altered. Customs officials, police and judges have always been part of the enforcement process, which encompasses a variety of interrelated activities. However, the substitution of these actors by private operators gives rise to the concentration of functions in the hands of a single actor. While this may certainly enhance efficacy and may be the adequate response to the volume of current trade, it also raises concerns from the point of view of independence, accountability¹⁹ and fair trial standards.

A second and related phenomenon concerns automation – that is, the fact that machines themselves implement the law. Computers use machine learning techniques to derive legal consequences, implement orders and reach conclusions from a database of primary sources.²⁰ In addition to automated processes that apply to all branches of the law, there are sui generis types of automation applicable to IP. For instance, machines may respond to copyright-based Internet takedown requests, and may also determine the existence of trademark infringement thanks to image recognition.²¹ The changes brought about by AI do not merely impact the online environment but also the manner in which judges adjudicate IP cases, lawyers practice IP law and authorities, including customs authorities, identify infringement. While due process concerns arise,²² many start-ups have developed applications that are transfiguring law practice,²³ in particular those that enable the automation of time-consuming processes, such as conducting research, writing memos, undertaking due diligence and collecting evidence.²⁴ Automation of the law is promoted invoking practical advances in terms of efficiency, savings, consistency in applying legal doctrines and legal harmonization.²⁵

A third change relates to centralization. Under this term, reference is made to the fact that a single action – writing a bit of software code – produces legal decisions for many individuals at once.²⁶ These are decisions that have previously required an individualized procedure.²⁷ Up until recent times, enforcement was human-driven, dynamic, public and individualized. Public authorities enforcing intellectual property – judges, customs officials and police – were ‘just’ humans. Similarly, intellectual property could not be understood as an automated process, but as a process of weighing and balancing competing rights and interests,²⁸ a process taking place in an individualized fashion in public settings and administered either by judicial or administrative authorities. These features contrast with a new situation where enforcement mechanisms are automated, centralized and privatized, and big (private) players occupy the place of public authorities.

C Digitalization and Use of Big Data in Customs Control

I Digitalization and Customs Control

While international logistics value chains are still characterized by the abundance of manual and paper-based processes, digitalization and AI are already transforming customs control. For instance, in the area of migration control, international travellers face, more and more often, semi-automated border controls that combine biometric screening,²⁹ facial recognition, automated lie detection and predictive modelling.³⁰ Likewise, predictive analytics also transform and enhance customs risk management regarding both the control of people and the monitoring of goods.

AI functional applications, such as computer vision, natural language processing and predictive analytics, are powered by the vast amount of data available in digitalized forms. Given the importance of data, a previous step is the digitalization of customs operations, an area that is characterized by red tape³¹ and the existence of a vast amount of valuable information transmitted in each transaction.³²

In the European Union, digitalization of customs was initiated back in the late 1990s³³ and dynamically promoted in 2008 with the adoption of the decision on a paperless environment for customs and trade.³⁴ The ‘e-Customs Decision’ identified the objectives, means and framework for setting up an electronic environment for customs and trade,³⁵ and was followed by the electronic customs Multi-annual Strategic Plan.³⁶ Since 2016, the Union Customs Code sets the framework on the rules and procedures for customs in the European Union.³⁷ Crucially, it mandates that ‘all exchanges of information … as required under the customs legislation, shall be made using electronic data-processing techniques’ and identifies 2020 as the deadline for a paperless customs union.³⁸

The compromise to move towards digitalized customs management has been extended to European Union trade partners via free trade agreements. The border enforcement sections of the intellectual property chapters of those trade agreements include compromises and best efforts–type of provisions to adopt digital management systems to monitor customs procedures. For instance, in the border enforcement sections of the 2019 EU association agreements with Japan and Vietnam, the use of electronic systems for the management of customs by applications of IP holders is encouraged, and the use of risk analysis to identify goods suspected of infringing IPRs is mandated.³⁹

II Big Data, Customs Control and Risk Analysis

In the area of customs enforcement, big data provides new knowledge, drives value creation, fosters new processes and enhances well-informed decision making.⁴⁰ Reference is commonly made to three characteristics of big data:⁴¹ high-volume, high-speed and high-variety of information assets.⁴² The fact that big data is made up of ‘extremely large data sets’⁴³ echoes another, implicit, characteristic: most often, big data is made of ‘raw’ information of both public⁴⁴ and private nature which is incomplete and imperfect.

Big data enhances the potential of descriptive, predictive and prescriptive data analytics, namely the obtention of raw data and the examination of that information with the purpose of identifying patterns, drawing conclusions, predicting the future and providing the best solution.⁴⁵ Regarding risk management in the area of customs enforcement, big data allows to predict threats, monitor trends and target high-risk transactions. Nowadays, both descriptive risk rating and prescriptive risk management have been incorporated into customs enforcement operations. This is where text analytics, data mining, statistics, natural language processing and machine learning offer valuable information and display patterns of infringing activities.

Automation and learning from past events enhance the efficiency of selecting risky cargos, in particular by identifying suspicious networks and transactions. Automation also facilitates the estimation of potential loses and damages. Additionally, image recognition is of relevance to trademark counterfeiting. In this regard, analysis and exploitation of images for automatic verification of consistency against available information is instrumental to perform customs risk management intended to fight counterfeiting.⁴⁶ Devices and software developed to spot fakes in ordinary places of commerce, such as shops,⁴⁷ can be adapted to conduct similar preliminary assessments by customs authorities.

The question that arises is how to take bigger advantage of currently existing technical capacities and amounts of data. Three aspects seem critical: First, it is necessary to ensure the quality of data regarding cargos, shipments and conveyances, and also relevant information concerning intellectual property rights. Incompleteness and heterogeneous data formats⁴⁸ make it more difficult to efficiently use information for analytical purposes.⁴⁹ Second, it is also important to widen the scope of the data used for analytical purposes. In this last respect, a possibility is to go beyond data of purely customs nature and correlate such data with, for instance, tax, crime or IP-related information.⁵⁰ Third, it is also necessary to overcome country differences in terms of capacity to implement common risk criteria and standards,⁵¹ and to efficiently address IP infringement having a border enforcement component.

III Distributed Ledger Technologies and Localization of Traded Goods

Distributed ledger technologies allow moving from a single administrator who controls the ledger where information is stored to a ledger shared by a network of stakeholders. None of the members of the network has autonomous control and all changes made by a member of the network are visible and transparent.⁵² A well-known example of public distributed ledger technology⁵³ is blockchain, which enables the recording of transactions between parties in a secure and permanent way while removing intermediaries that previously verified transactions – a function that is of relevance to many economic and technological fields.⁵⁴

Efficiency, velocity, transparency, traceability and automation are the main advantages brought by distributed ledger technologies to international logistics and commercial processes.⁵⁵ Distributed ledger technologies significantly reduce bureaucracy and paperwork,⁵⁶ and may also enable new business models⁵⁷ and enhance asset management.⁵⁸ They furthermore make it possible to obtain information in respect of manufacturing practices, quality attributes and place of origin. In the area of intellectual property, traceability is among the most visible advantages brought by distributed ledger technologies, in particular for goods protected by IPRs that identify the origin of goods (geographical indications) and for goods protected by IPRs that enable consumers to identify specific producers, characteristics and qualities (trademarks).

A particularly valuable application in the area of customs control is the possibility offered by the Internet of Things (IoT) to capture the location, condition and status of traded goods in real time.⁵⁹ This allows logistics service operators to detect irregularities affecting the cargos, improve supply chain control and detect the introduction of fake products. In order to take full advantage of distributed ledger technologies, track and trace methods should ideally allow interaction, which means that some physical devices may be embedded in the traded goods that are object of control.

Track and trace methods belong to the broader group of anti-counterfeiting technologies, which include overt and covert authentication technologies that determine whether a product is original, and allow for enhanced control within the supply chain. Regarding authentication technologies, overt technologies are accessible using human senses, such as vision and touch, thus they do not need any specific physical device,⁶⁰ while covert technologies are hidden and only accessible to technology providers, brand owners or authorized stakeholders.⁶¹ In respect of track and trace technologies, optical methods and Radio Frequency Identification (RFID) tags stand out.⁶² Both of them allow for the localization of the product along the production and distribution chain: optical technologies consist of a code that contains information on the product and is affixed on the product itself,⁶³ whereas RFID tags can be read by radio waves.⁶⁴ As it has been noted, ‘smart devices can be securely tied to or embedded in the physical product to autonomously record and transmit data about item condition including temperature variation, to ensure product integrity, as well as any evidence of product tampering’.⁶⁵ In the area of pharmaceutical products, for instance, this allows to asses at the same time several types of IP infringement and the compliance with regulatory standards.

D Legal Challenges and Opportunities

The control of intellectual property protected goods in transit⁶⁶ exemplifies well how existing legal challenges can be addressed resorting to AI applications and big data.

I Opportunities

TRIPS does not require the control of goods protected by IPRs in transit, nor the monitoring of exports. It just orders impeding the importation of counterfeit and pirated goods.⁶⁷ Building on this minimum standard, a significant number of countries has enacted legislation that goes beyond such protection,⁶⁸ thus it is nowadays usual to find countries that control exports and, to a lesser extent, the transit of IP protected goods. This is particularly the case with trademark and copyright protected goods, but some countries also monitor goods protected by other IP categories.⁶⁹

In the case of the control of goods in transit, the goal of the country of transit is to impede the arrival of products to foreign jurisdictions. This may generate tensions with two fundamental principles – the principle of territoriality of IP protection and the principle of freedom of transit of internationally traded goods. Such tension is severe in respect of the principle of territoriality, since some countries have put in place customs controls of goods in transit that disregard the status of IP protection in the country of origin and in the country of destination. In these cases, seizure and eventual destruction are decided according to the law of the country of transit, hence eroding the principle of territoriality. In the case of the principle of freedom of transit, the tension is also acute, and both legal reform and courts have qualified the cases where restriction of freedom of transit is acceptable to control goods protected by IPRs. The legal standards developed are intended to preserve the balance between adequate IP protection and the principles just referred to. As argued earlier, however, they could benefit from technological advances in the area of big data and AI.

Pursuant to EU Regulation No 608/2013, goods are suspected of infringing an intellectual property right if there are indications that where such goods are found in a member state, they are the subject of an act infringing an intellectual property right.⁷⁰ However, the mere transit of goods through a country where they are protected does not imply an infringement.⁷¹ In response to the heated debates arising from the detention of medicines in transit,⁷² EU Regulation No 608/2013 stated that ‘customs authorities should, when assessing a risk of infringement of intellectual property rights, take account of any substantial likelihood of diversion of such medicines onto the market of the Union’.⁷³ The Court of Justice of the European Union (CJEU) linked the detention and the suspension of release of protected goods in transit to the potential diversion of those goods onto the transited market, thus only by placing the goods in the internal market can the subject matter of a specific intellectual property right be infringed.⁷⁴ Likewise, if goods are the subject of a commercial act directed to European consumers, intellectual property rights may be infringed and goods placed under a suspensive procedure may be detained. The risk of fraudulent diversion to European consumers may also arise in other circumstances, even when goods have not yet been directed towards European consumers. Customs authorities can, in effect, detain the goods or suspend their release when there are indications that commercial activities may take place in the near future or are being disguised.⁷⁵ Suspicion, based on a number of facts of the case, will suffice for that purpose. The CJEU gave a number of examples, including the destination of the goods not being declared, the lack of precise or reliable information as to the identity or address of the manufacturer or consignor of the goods, a lack of cooperation with the customs authorities, or the discovery of documents suggesting that there is a likelihood of a diversion of those goods to EU consumers.⁷⁶

The control of the scenarios described in the paragraph above is difficult in the daily operation of customs control. Some of the examples given by the CJEU require that customs officials and/or right holders have access to information that, most often, is not publicly available. On other occasions, the volume of internationally traded goods makes it almost impossible to manage existing information. These are problems, however, that big data and AI can help to overcome. In the first case, if the use of distributed ledger technologies and track and trace technologies became general, the efficiency of control would increase exponentially. As explained by Okazaki, advanced sensor technology ‘allows logistics service providers to detect any irregularities occurring in or around the cargos in transit, thus helping to enhance supply chain security. As such, a containerized cargo being once regarded as “low-risk” or “risk-free” can maintain the same condition until it is delivered at the destination unless any suspicious intervention is detected during the time of transport’.⁷⁷

Management of data at the international level can also be instrumental in another, very practical, context. In order to protect IP, some countries have introduced legal regimes that, while allowing to take action regarding goods in transit, still differ from the EU model described earlier. An interesting, while controversial, alternative consists of anticipating the moment and location of the protection. This is the model followed by Switzerland, where the patent owner can impede the transit of patent-infringing goods if he can also prohibit the import into the country of destination.⁷⁸ Hence, Swiss law permits anticipating the moment of the protection that the same title holder could demand in the country of destination. Although it relates to patents, Swiss authorities justify such anticipation ‘in view of the increasing international dimension of counterfeiting and piracy’,⁷⁹ and in order to ‘prevent Switzerland from becoming a transit country for pirated goods’.⁸⁰ In the context of trademark law, the EU has also made relevant the law of the final country of destination. In the EU, the entitlement of the trademark proprietor to detain products in transit shall lapse if, in the context of customs procedures, ‘evidence is provided by the declarant or the holder of the goods that the proprietor of the registered trade mark is not entitled to prohibit the placing of the goods on the market in the country of final destination’.⁸¹

Naturally, should border authorities have direct and speedy access to the database of intellectual property offices from all over the world, they could also rapidly verify whether the IP owner who claims to have the right to impede importation to the country of destination is entitled to do so. It is difficult to envisage such a system to function ex officio, but it would accelerate procedures if the process were initiated at the request of an interested party. The same applies in respect to the holding made by the proprietor or consignor of the goods in the example provided regarding the EU. Advanced analytics allows to correlate internal data with other categories of data. In particular, it allows to correlate the customs situation, national intellectual property and international intellectual property protection status in the country of destination. Thus, putting existing sources of information and technologies at work for the benefit of customs authorities would make it more feasible to meet standards that, right now, are rather difficult to attain because of technical and resource-related constraints.

II Challenges

While big data and AI bring about new opportunities in the context of IP enforcement, including customs enforcement, new challenges also emerge. On this occasion, concerns arising from automation of the law and due process restrictions become also of relevance in respect of border enforcement.

In contrast to the TRIPS standard of releasing the goods that have been detained while in transit, the EU has put in place a speedy process for the destruction of goods suspected – but not confirmed – of infringing IPRs. TRIPS mandates the release of goods in case proceedings leading to a decision on the merits of the case have not been initialed or provisional measures have not been adopted within a period of ten working days after the applicant has been served notice of the suspension.⁸² However, the EU has inverted the logic behind this rule and has established that the destruction of the goods will follow, without any further procedure on the merits of the case, if the alleged infringer does not respond in due time to the seizure.⁸³

It is predictable that the automation of procedures will result in more cargos being detained and more notifications of such detentions being sent to the owners of the cargos. However, economic and operational difficulties to respond to this type of processes, taking place in different continents and eventually exceeding what small companies can afford, will persist. In many instances, the owner of the goods may not contest the detention because, for instance, doing so may be more expensive than the value of the parcel that has been detained, or just for lack of knowledge or lack of time to react. If other, compensatory, measures – also of a technological nature – are not introduced, due process standards, in particular the right to a fair hearing and the presumption of innocence, become clearly threatened, especially when the right holder does not even need to start procedures on the merits of the case.

Reflections made by Citron in respect of due process and algorithmic enforcement, and the need to ensure that analytical algorithms satisfy standards of review guaranteeing fairness and accuracy, are fully applicable to customs enforcement.⁸⁴ A number of actions would mitigate those concerns. First, it is necessary to improve transparency (or at least the understanding) of the algorithms that determine which cargos and goods will be detained and inspected.⁸⁵ Next, it is also necessary to allow challenging the decision and detention undertaken with the assistance of automated mechanisms. Finally, it must also be possible to enable public oversight of automated border enforcement.⁸⁶ In reality, these concerns are not really different from those expressed in other areas of IP enforcement where automation has become a common practice, as discussed earlier.

E Conclusions

Digitalization, big data and distributed ledger technologies drastically change law enforcement. When applied to customs control, as it happens in other domains, these technologies result in cost savings and promote more efficient and less-prone-to-error administrative, judicial and commercial processes. Interconnectedness, instant access to foreign databases and constant monitoring of the precise location of goods allow to implement, for instance, standards relating to the control of goods in transit that were difficult to meet until now. Similarly, AI functional applications, such as image recognition, combined with the possibility to constantly and exponentially learn from past events, strengthen systems to control internationally traded goods protected by intellectual property rights.

Together with opportunities, challenges of both technical and legal nature also arise. Technical challenges are still manifold and relate to aspects such as the low quality and heterogeneous formats of digitalized data feeding AI functional applications. Legal concerns expressed in respect of algorithmic law enforcement relate to transparency, accountability and contestability of decisions. These concerns, which are of a general nature, are also of relevance to customs enforcement, as seen in the case of the automation of decisions concerning goods in transit. Algorithmic law enforcement must respond to the mentioned challenges and acknowledge in particular that intellectual property enforcement is a process of weighing and balancing rights and interests of different nature, and not an automated process to implement a predefined decision. Discretion and proportionality are central in the enforcement process, but these are attributes of remedies that require human virtues and skills that the current level of technological development does not seem capable to replicate yet.