One step closer towards a legal framework for artificial intelligence – European Parliament Resolutions 20 October 2020

Introduction

Artificial intelligence (AI) surrounds us from all sides: from cell phones to all kinds of online services to autonomous cars. AI and products equipped with it have a large number of supporters, but also people who look at the created solutions carefully, sometimes with concern. It is hardly surprising when we hear about yet another accident caused by an autopilot or chatbots spreading hate speech. The topic of the principles of responsibility for the creation of artificial intelligence (equipping the AI with certain functions and giving it appropriate learning capabilities), but also for the functioning of artificial responsibility, has long been present not only in discussions among lawyers but also particularly vivid among the public. The dilemmas and aspects to be taken into account are indeed many. All the more reason to welcome the information about the next steps towards the regulation of AI issues. Three resolutions adopted by the European Parliament (EP) provide a basis for future legislation that will address the problems that have arisen around AI liability:

1. Resolution of 20 October 2020 with recommendations to the Commission on a framework for the ethical aspects of artificial intelligence, robotics, and related technologies (2020/2012(INL));
2. Resolution of 20 October 2020 with recommendations to the Commission on a civil liability regime for artificial intelligence (2020/2014(INL));
3. Resolution of 20 October 2020 on intellectual property rights in the field of artificial intelligence technology development (2020/2015(INI)).

As can be seen from the above summary, the resolutions indicated above deal with three aspects: ethical aspects of AI, civil liability for AI, intellectual property rights vs. AI. The following post, in line with the theme of the resolutions, will be divided into three parts and will be a selection of the more interesting issues.

Ethics of AI

According to the resolution on the framework for the ethical aspects of artificial intelligence[1], any legislative action about artificial intelligence, robotics, and related technologies should comply with the principles of necessity and proportionality to enable companies to bring innovative products to market. Importantly, as AI-based solutions require access to vast amounts of data, the EP endorses the Commission’s proposal to create a common data space, which will ensure that data protection laws are respected. The development and implementation of AI itself must always respect ethical principles and take into account human leadership and democratic oversight.

The EP Resolution stresses that high risk IS should be singled out (e.g., by recognizing that there is a significant risk of causing injury or harm to individuals due to the use of the technology). The risk assessment should be based on a comprehensive and cumulative list of high-risk sectors and high-risk uses and purposes.

In the resolution, we can find a division into sectors, which are important from the perspective of its subject matter. Thus, the EP points out that the consumers’ right to information is one of the key rights in the EU, and therefore it should also apply to the IS (in particular, it should include transparency of interactions with the IS systems and how they function). Another also highlights that AI has the potential to create and reinforce unequal treatment, as well as to create forms of automated discrimination. Therefore, any individual or legal entity should have the right to seek redress for a decision made by an AI.

One of the most relevant issues is the section on privacy and biometrics. Of course, special care is taken to ensure that data processing using AI complies with the RODO. We also find here guidance for public sector entities, e.g. public institutions using remote recognition technologies (e.g. facial recognition) should disclose this fact and their use should always be proportionate, targeted and limited to specific purposes, limited in time, and carried out by Union law.

An equally important issue, which is mentioned on the pages of the Resolution, is the proposal of a European ethical certification system. It would be obtained at the request of any entity developing, implementing, or using technologies not considered high-risk that wishes to obtain certification of a positive conformity assessment by the relevant national supervisory authority. This type of certification, according to the EP, would encourage the consideration of ethics at the design stage throughout the supply chain of artificial intelligence ecosystems.

Civil liability of AI

The Resolution on Civil Liability for Artificial Intelligence[2] emphasizes that one of the most important issues in political action today is the implementation of AI systems in society, the workplace, and the economy. This broad framing of the topic only indicates that the more widely artificial intelligence surrounds us, the more necessary it is to regulate civil liability for damages caused by it. The Resolution notes that “all physical or virtual activities relying on, devices or processes using AI systems can in principle be a direct or indirect cause of harm, while they are almost always the result of someone having constructed, implemented or interfered with such a system (…).” Interestingly, the EP believes that it is not necessary to give AI systems legal personality (which is also repeatedly raised in discussions as a way to address this issue).

The resolution indicates that the person responsible for damages caused by an IS operation should be the operator of the IS system (a term that would include the front-end and back-end operator) since in many cases he or she will be the first identifiable contact person for the victim. The type of AI system over which the operator has custody would have an element indicating responsibility. For this reason, the EP suggests introducing a strict liability regime for autonomous AI systems that are high risk. This type of AI would be in an annex to the regulation, which would be reviewed every 6 months and updated as necessary. It is worth noting that strict liability would have to be accepted for the IS systems covered by the Annex, while in other cases liability based on the fault is sufficient. For this reason, operators of such systems should have civil liability insurance. Nevertheless, an operator will be able to avoid liability based on a fault if it can prove that due care has been taken. To this end, manufacturers should be required to appoint an AI liability representative within the EU as a point of contact to answer any questions from operators.

It is also worth mentioning that it is recommended that in exceptional cases (e.g. in the case of collective damages exceeding the maximum amounts of compensation that will be specified in the regulation) special compensation funds should be established for a certain period to cover damages.

The draft regulation, which is attached to the Resolution, also includes a proposed statute of limitations for SI liability claims of 10 years from the date of property damage or verifiable economic loss resulting from serious intangible harm, or 30 years from the date of a high-risk AI operation that destroyed the victim’s property or serious intangible harm.

Intellectual property rights and AI

In its resolution on intellectual property rights in the field of the development of artificial intelligence technologies[3], the EP points out that the significant progress in the development of AI can raise serious problems, e.g. in the form of making it difficult to distinguish between objects of intellectual property rights created by humans and those created by AI, which consequently makes it impossible to fairly reward “human authors” whose original works are used to drive such technologies, regulation in the form of a regulation should be introduced to avoid fragmentation of the European Digital Single Market. In particular, concerning the problem presented above, it is important that if an author uses the AI as a mere enabler, then the existing regulations on intellectual property are sufficient to protect such a creation. At the same time, the Resolution stresses that creations based solely on the IS should be protected under intellectual property rights, which, however, in some cases may be problematic due to the principle of originality related to the individual and the fact that “intellectual creation” refers to the personality of the author.

Another issue to which the EP draws attention is the fact that AI and related technologies are based on computational models and algorithms that are considered to be mathematical methods and are therefore not patentable as such, and for this reason, it is worth considering the impact of such potential patent protection.

To best address this issue, the EP recommends that AI be captured in neutral definitions that do not become obsolete with technological progress.

In the context of IPRs in the light of AI technology, the degree of human intervention, the autonomy of the system itself, or the importance of the role and origin of the data used should be taken into account.

What’s in the future?

The year 2020 is extremely fruitful in terms of proposals to regulate AI issues. As a reminder, we should mention, among others, the “White Paper on Artificial Intelligence” or the “Conclusions on the Charter of Fundamental Rights in the context of Artificial Intelligence and Digital Change” (which we also wrote about on IP blog and raised these issues in the podcast – I refer to these materials and encourage you to read them). It is to be welcomed that the above resolutions, together with the attached draft regulations, undoubtedly indicate the direction in which the European Union wants to go when regulating artificial intelligence, as evidenced by concrete proposals put forward by the European Parliament, such as accepting strict liability for high-risk AI technologies. The resolutions and other proposed regulations discussed in this text are all the more reason to keep an eye on the legislative process of implementing regulations on artificial intelligence at the EU, and later at the national level.

Documents:

1. Resolution of 20 October 2020 with recommendations to the Commission on a framework for the ethical aspects of artificial intelligence, robotics and related technologies (2020/2012(INL)) – https://www.europarl.europa.eu/doceo/document/TA-9-2020-0275_PL.pdf;
2. Resolution of 20 October 2020 with recommendations to the Commission on a civil liability regime for artificial intelligence (2020/2014(INL)) – https://www.europarl.europa.eu/doceo/document/TA-9-2020-0276_PL.pdf;
3.  Resolution of 20 October 2020 on intellectual property rights in the field of artificial intelligence technology development (2020/2015(INI)) – https://www.europarl.europa.eu/doceo/document/TA-9-2020-0277_PL.pdf.

[1] Resolution of 20 October 2020 with recommendations to the Commission on the framework for the ethical aspects of artificial intelligence, robotics, and related technologies (2020/2012(INL))

[2] Resolution of 20 October 2020 with recommendations to the Commission on a civil liability regime for artificial intelligence (2020/2014(INL)).

[3] Resolution of 20 October 2020 on intellectual property rights in the development of artificial intelligence technologies (2020/2015(INI)).