Books banner

1. Introduction

Welcome to this module on the legal and ethical aspects of data and AI development and use. In this module, you will be introduced to the most relevant and important issues currently affecting how the AI industry in the EU – and international organisations working with EU companies – carries out research and development. This module is set out in three parts, corresponding with the three scheduled self-directed study days, guiding you through the learning content, and helping you to check your understanding of the information.

The aims of this module are:

  • Explaining the legislative or law-making process in the EU, and current EU legislation that affects the process of AI development and use.

  • Explaining the EU's approach to AI and the reasons for the approach.

  • Examining the current EU legislation and rules concerning development and use of AI.

  • Examining the potential legislation that will directly impact how AI development and use will be regulated in the EU.

  • Analysing the framework put into place by the European Commission for the industry to use in promoting and encouraging the ethical development and use of AI.

  • Identifying the persons and organisations responsible for ensuring that the rules and requirements are enforced, and who provide guidance and information to the EU AI industry.

At the end of the module, you should be able to:

  • Describe the process of law-making in the EU.

  • Identify the legislation and rules put in place for AI in the EU and understand their function and purpose.

  • Recall the most important of these rules and principles to bear in mind when engaged in an AI system development process.

  • Identify the available methods to self-assess for legal and ethical compliance in developing AI systems.

  • Integrate these self-assessment methods with ongoing AI system development processes, and identify issues to resolve, or seek appropriate expert or official guidance for if necessary.

  • Identify the available resources for information, advice, guidance, and best practices for AI system development and use.

Make sure to upload all the answers to the questions listed below to your GitHub account.

2. The Proposed Artificial Intelligence Act

In April this year, the European Commission unveiled the Artificial Intelligence Act (AI Act). At the time of writing, the proposed AI Act is going through the approval process by all 27 EU member states. When approved, it will be the first legislative and governance framework to comprehensively address the creation, deployment, and use of artificial intelligence systems in the EU. A fundamental aspect of this framework is data protection and privacy, among other human rights recognised in the EU. Similar to the GDPR, it is intended for this proposed legislation to have extra-territorial effect. It is anticipated that it will come in to force sometime in 2022/3. The basis of the AI Act is the EU's underlying notion that developing trustworthy technologies will benefit the global acceptance of AI in Europe. What is important for the European Commission is building trust, and fundamental to this is requiring the appropriate protection of people's fundamental rights and promoting safety. This goal of trust-building can be attained by drawing boundaries around the why and how of developing and using AI systems. The challenge here is then ensuring that these boundaries are not so onerous that they hinder the creation of AI solutions. How does the AI Act attempt to manage this difficult balancing act between regulation and competitive innovation? The drafters of the proposed AI Act have tried to find a middle way by using an approach that has a scale of regulation that is attuned to the level of risk of an AI system:

  • where it concerns what are considered unacceptable uses of AI, these are banned outright;

  • where it concerns uses that are important but carry a high risk, these are regulated stringently;

  • where it concerns uses that have a limited risk or no obvious risk; it encourages the use of industry codes of conduct.

The message that the EU is sending is essentially that it champion and enforce AI that is trustworthy, and that it will pursue an agenda of "responsible competition" in terms of its AI goals globally. Many organisations within the EU, and even those outside the EU, especially those from countries that have economic and research links with EU states, are studying these proposed rules carefully, and are preparing to comply with them.

2a Read the articles linked below for an explanation of the key aspects of the proposed AI Act. The first linked article has a helpful visual guide and also suggests, with a process flow diagram, how go about thinking and acting when contemplating the creation of a high-risk AI system (which can be scaled for lower-risk AI systems, too).

  • Read an article about A (more) visual guide to the proposed EU Artificial Intelligence Act by Lukianets, N. (2021, May 5) by clicking this link.

  • Read an article about The Artificial Intelligence Act by Mueller, B. (2021, May 4) by clicking this link.

2b Watch this webinar recording discussing the EU's approach to artificial intelligence – the discussion also ventures into how governments are around the world are engaging with the EU on this proposed legislation.

After you have watched the above video, reflect on what the key takeaways are for you, and write them down. Share these takeaways with your classmates.

2c Read Articles 1 – 24 of the proposed AI Act, using the guide tips in the above readings to help you navigate it. You will see that the document has three parts:

  • The Explanatory Memorandum (Parts 1 - 5)

  • The Proposal (Paragraphs 1 – 89)

  • The Titles and Articles

For the purposes of this module, focus on Articles 1 – 24.

2d Check your understanding:

  • Which Article of the AI Act provides definitions of the terms used in the Act?

  • Is the list of artificial intelligence processes and approaches referred to in Article 3 and listed in Annex 1 of the AI Act a closed or open list? Which other Article addresses this issue, given that technological progress is always happening?

  • For each of the categories of AI systems that are identified in the AI Act according to risk, give one example of an actual AI system (still in use or no longer in use) that matches the risk level.

Literature

Artificial Intelligence Act: What Is the European Approach for AI? (2021, October 16). Lawfare. Retrieved December 2, 2021, from https://www.lawfareblog.com/artificial-intelligence-act-what-european-approach-ai

Lukianets, N. (2021, May 5). A (more) visual guide to the proposed EU Artificial Intelligence Act. Medium. Retrieved December 2, 2021, from https://lukianets.medium.com/a-more-visual-guide-to-the-proposed-eu-artificial-intelligence-act-64ae0e550ac3

Mueller, B. (2021, May 4). The Artificial Intelligence Act: A Quick Explainer. Center for Data Innovation. Retrieved December 2, 2021, from https://datainnovation.org/2021/05/the-artificial-intelligence-act-a-quick-explainer/