Part III: Advanced Topics in AI and Data Protection
By the end of this part, learners will be able to:
- design assessment practices to evaluate the technical aspects of AI systems and their operation within an organization.
- propose technical and organizational measures to ensure that data subject rights are properly addressed throughout an AI system’s life cycle.
- devise different information disclosure strategies in line with the legal requirements directed at each type of information recipient; and
- evaluate various sources of guidance that can support organizations in specifying their legal duties.
So far, this book has offered an overview of the life cycle of AI models and systems within an organization. Each of the stages covered in Part II can give origin to various risks to the protection of personal data, which a data protection professional must address. To understand what is unique about AI in those contexts, that professional requires an understanding of the technical side of AI technologies and the vocabulary to dialogue with technical and business stakeholders, tasks that are supported by the contents of Part I of this module. Now, the remaining five chapters will cover specific issues that are likely to span more than one stage of the life cycle.
It would not be feasible to offer an exhaustive coverage of all such issues. Each AI system or model undergoes its own life cycle, and processes that are central for a specific project might play a minor role in another. Still, current experiences with the design and implementation of AI systems and models suggest that some problems appear more often than others. This means the selection of issues for this part is driven by two main concerns:
- The issues covered by each chapter will be relevant for most, if not all, AI systems developed or deployed within the EU.
- Approaches to those issues offer insights that can be used to tackle other issues related to data protection in AI systems.
The knowledge and skills developed in this Part should, therefore, be applicable to a broad range of solutions, even when the specific solutions proposed here are not.
To this effect, Part III is formed by four chapters:
- In 10 Fairness and Accountability for AI, learners will engage with two kinds of assessments of AI systems: technical audits and impact assessments.
- 11 Transparency towards Stakeholders focuses on the information disclosure duties that organizations have with regard data subjects and public authorities, including the so-called right to an explanation.
- 12 Regulating AI by Design will unpack the concept of “regulation by design”, show how it might contribute to compliance, and raise some warnings about the limits of technical interventions in promoting data protection principles.
- Finally, 13 Supporting the Lawful Use of AI discusses how organizations can evaluate whether non-binding sources, such as technical standards and certification schemes, offer sufficient guidance for their compliance needs.
Mastering those topics will help learners in facing problems that cannot be confined to a specific stage of the AI life cycle.