Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

The EU AI Act is a good start but falls short

Chalisa Veesommai Sillberg, Jose Siqueira De Cerqueira, Pekka Sillberg, Kai-Kristian Kemell, Pekka Abrahamsson

TL;DR

The paper investigates how enterprises can comply with the EU AI Act by aggregating industry and academic perspectives through a Multivocal Literature Review (MLR) and NLP analysis. Using 130 sources (56 included) and 26 gray literature items, it applies TF-IDF and Latent Dirichlet Allocation to identify three principal challenges—liability, discrimination, and tool adequacy—and three core strategic directions: risk-based regulatory compliance, ethical frameworks in development, and regulatory risk management. The study validates its findings against existing literature, offering a practical framework to minimize regulatory distance to the EU market and guiding governance practices across high- and low-risk AI deployments. It also highlights negative sentiments around regulatory interpretations and transparency, suggesting sector-specific future work, such as deep dives into healthcare, finance, and transportation to tailor compliance guidance.

Abstract

The EU AI Act was created to ensure ethical and safe Artificial Intelligence (AI) development and deployment across the EU. This study aims to identify key challenges and strategies for helping enterprises focus on resources effectively. To achieve this aim, we conducted a Multivocal Literature Review (MLR) to explore the sentiments of both the industry and the academia. From 130 articles, 56 met the criteria. Our key findings are three-fold. First, liability. Second, discrimination. Third, tool adequacy. Additionally, some negative sentiments were expressed by industry and academia regarding regulatory interpretations, specific requirements, and transparency issues. Next, our findings are three essential themes for enterprises. First, risk-based regulatory compliance. Second, ethical frameworks and principles in technology development. Third, policies and systems for regulatory risk management. These results identify the key challenges and strategies and provide less commonly discussed themes, enabling enterprises to align with the requirements and minimize their distance from the EU market.

The EU AI Act is a good start but falls short

TL;DR

The paper investigates how enterprises can comply with the EU AI Act by aggregating industry and academic perspectives through a Multivocal Literature Review (MLR) and NLP analysis. Using 130 sources (56 included) and 26 gray literature items, it applies TF-IDF and Latent Dirichlet Allocation to identify three principal challenges—liability, discrimination, and tool adequacy—and three core strategic directions: risk-based regulatory compliance, ethical frameworks in development, and regulatory risk management. The study validates its findings against existing literature, offering a practical framework to minimize regulatory distance to the EU market and guiding governance practices across high- and low-risk AI deployments. It also highlights negative sentiments around regulatory interpretations and transparency, suggesting sector-specific future work, such as deep dives into healthcare, finance, and transportation to tailor compliance guidance.

Abstract

The EU AI Act was created to ensure ethical and safe Artificial Intelligence (AI) development and deployment across the EU. This study aims to identify key challenges and strategies for helping enterprises focus on resources effectively. To achieve this aim, we conducted a Multivocal Literature Review (MLR) to explore the sentiments of both the industry and the academia. From 130 articles, 56 met the criteria. Our key findings are three-fold. First, liability. Second, discrimination. Third, tool adequacy. Additionally, some negative sentiments were expressed by industry and academia regarding regulatory interpretations, specific requirements, and transparency issues. Next, our findings are three essential themes for enterprises. First, risk-based regulatory compliance. Second, ethical frameworks and principles in technology development. Third, policies and systems for regulatory risk management. These results identify the key challenges and strategies and provide less commonly discussed themes, enabling enterprises to align with the requirements and minimize their distance from the EU market.

Paper Structure

This paper contains 12 sections, 5 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Search strategy and literature review protocol
  4. Quality assessment for the literature review approach
  5. Data analysis approach
  6. Results
  7. MLR execution, protocol and quality assessment
  8. Interpreted insights from the Data Analysis Approach
  9. Discussion
  10. Key challenges
  11. Themes for strategy development
  12. Conclusion and Future Work

Figures (5)

  • Figure 1: The system architecture of this study.
  • Figure 2: The results of a) the selected data sources from pre-selection and b) the results of data source quality assessment.
  • Figure 3: The extracted results of a) the challenges and b) the strategies and processes required in complying with the EU AIA.
  • Figure 4: The framework outline of key challenges and strategies when complying with the EU AIA.
  • Figure :