AI Regulations: Innovation and Safety in EU's Landscape

The European Parliament has made significant progress towards ethical AI development, endorsing transparency and risk-management rules for AI systems. The MEPs have amended the Commission’s proposal to establish a technology-neutral AI definition, applicable to present and future systems.

AI regulations in EU

Grand Thought Leadership

Discover the world of compliance with Grand's captivating articles. In this easy-to-understand series, we dive into recent news from trusted compliance sources, bringing you intriguing insights, timely regulatory updates, and helpful expert views. Stay ahead in the ever-changing world of compliance and face challenges with confidence.

Start your Journey Now

The Members of the European Parliament (MEPs) have taken a significant stride towards a human-centric and ethical development of Artificial Intelligence (AI) in Europe, endorsing new transparency and risk-management rules for AI systems. This initiative, which saw the Internal Market Committee and the Civil Liberties Committee adopt a draft negotiating mandate on the first-ever rules for Artificial Intelligence, is a crucial milestone in ensuring that AI systems are safe, transparent, traceable, non-discriminatory, and environmentally friendly.

The MEPs' amendments to the Commission’s proposal aim to establish a uniform definition for AI that is technology-neutral, allowing it to apply to both current and future AI systems. This approach shows foresight in recognizing the pace at which AI technology evolves, ensuring that the legislation remains relevant as advancements continue to emerge.

EU AI Regulations: Key points

  1. MEPs aim to establish a technology-neutral, uniform definition for AI, ensuring that regulations apply to both current and future AI systems.
    The rules follow a risk-based approach, with obligations for AI providers and users varying depending on the potential risk level associated with each AI system.
  2. Certain high-risk AI practices will be prohibited, including the deployment of manipulative techniques, social scoring, and the use of certain biometric and emotion recognition systems.
  3. The MEPs expanded the classification of high-risk AI to include AI systems that could potentially harm people's health, safety, fundamental rights, or the environment.
  4. MEPs also included obligations for providers of foundation models, a new and rapidly evolving field in AI, ensuring that these providers guarantee the protection of fundamental rights, health and safety, the environment, and the rule of law.

Implications, Consequences, and Hurdles

  • Technological Challenges: AI is a rapidly evolving field, and some of the requirements set forth by the MEPs, such as the need for AI systems to be transparent and traceable, may be technically challenging to implement. Transparency in AI, often referred to as explainable AI, is an active area of research, and while progress is being made, it's not always possible to provide clear explanations for how complex AI models, like deep learning neural networks. Similarly, ensuring traceability in AI systems, which involves being able to track and understand the decision-making process within these systems, can also be difficult due to the complexity and often 'black-box' nature of some AI models. These technological hurdles will require ongoing research and innovation, and close collaboration between regulators, AI developers, and the broader tech industry.

  • Innovation vs Regulation: Striking a balance between the need for regulation and the desire to stimulate AI innovation will be a significant challenge. Over-regulation could potentially stifle innovation, hindering the development of new AI technologies and applications. For instance, strict regulations might discourage start-ups and smaller companies from developing AI solutions due to the high cost of regulatory compliance. On the other hand, under-regulation could lead to misuse of AI or the development of AI systems that pose risks to individuals or society. Finding the right balance will require careful consideration of the potential benefits of AI against its potential risks, and may involve ongoing adjustments to the regulatory framework.

  • Risk Assessment: Determining the risk level associated with each AI system could be a complex task. AI systems can be used in a wide variety of applications, from relatively low-risk applications like recommending movies or music, to high-risk applications like autonomous vehicles or medical diagnosis systems. There may be disagreements about what constitutes an "unacceptable level of risk", and different stakeholders (like AI developers, users, and those potentially affected by AI systems) may have different perspectives on this. Moreover, the risk associated with an AI system might change over time as the system learns and adapts, or as it's used in different contexts.

  • Enforcement: Monitoring compliance with these regulations across the vast AI industry will be a difficult task, particularly given the global nature of the tech industry. AI development often involves collaboration across different countries and jurisdictions, and AI systems can be used and accessed from anywhere in the world. This raises questions about how the rules will be enforced for non-European companies, or for European companies developing AI systems for use in other parts of the world. Furthermore, given the complexity of AI systems and the technical expertise required to understand them, ensuring that regulators have the resources and expertise to effectively monitor compliance could also be a challenge.

AI Act: Ripple Effects

Global AI Standards: As the first legislation of its kind worldwide, these new rules have the potential to set a precedent for global AI standards, potentially influencing other countries' AI regulations. The EU has been influential in setting worldwide standards in other areas of technology regulation before - the General Data Protection Regulation (GDPR), for example, has inspired similar privacy laws in other parts of the world. However, the impact on global AI standards will depend on a variety of factors, including how effectively these rules are implemented within the EU, how they are perceived by other nations, and how well they are able to accommodate the rapidly evolving nature of AI technology. It is also worth noting that different cultural, social, and political contexts could lead to different approaches to AI regulation in different parts of the world.

Trust in AI: Public trust in AI has been a significant barrier to the adoption of AI technologies. Concerns about transparency, bias, misuse of data, and the lack of human oversight have contributed to a certain level of scepticism and unease about AI among the general public. If successfully implemented, these new rules could help to alleviate some of these concerns by ensuring that AI systems are safe, transparent, and overseen by humans. This, in turn, could increase public trust in AI and facilitate greater adoption of AI technologies across various sectors. However, building trust in AI is a complex and ongoing task that will require not just effective regulation, but also efforts to educate the public about AI and to involve a broad range of stakeholders in decisions about AI development and use.

Protecting Citizens' Rights: The new law aims to strengthen citizens' rights in relation to AI, providing mechanisms for people to file complaints about AI systems and receive explanations of decisions made by high-risk AI systems. This represents an important step towards ensuring accountability in AI, and could help to protect individuals from potential harm caused by AI systems.
However, there are challenges to be addressed. For instance, providing meaningful explanations for decisions made by complex AI systems is a difficult task, given the often opaque nature of these systems. Furthermore, the effectiveness of the complaints process will depend on the accessibility and responsiveness of the system, and on the ability of regulators to take appropriate action in response to complaints.

AI Act Effects
AI Act Effects

Key Assessments and Probabilities

Assessment Probability Justification
The rules will significantly increase public trust in AI0.7The rules address many public concerns about AI, such as the lack of human oversight and potential misuse of biometric data. However, their effectiveness will depend on successful implementation.
The rules will influence global AI standards0.8As the first legislation of its kind, it's likely to influence other countries' approach to AI regulation, especially considering the EU's past influence on global standards (e.g., GDPR).
The rules will effectively prevent misuse of AI0.5The effectiveness of the rules in preventing misuse will depend on the robustness of enforcement mechanisms, which, given the complexity and global nature of the AI industry, might be a significant challenge.
The rules will be technically challenging to implement0.8The requirement for AI systems to be transparent and traceable, among other things, may pose significant technical challenges, given the current state of AI technology.
The rules will remain relevant as AI technology evolves0.7The proposed technology-neutral definition of AI is designed to apply to both current and future AI systems. This indicates an attempt to future-proof the regulations, but the rapid pace of AI development means that the effectiveness of this approach is uncertain.

AI regulation thoughts from Grand Compliance

  1. Open Source and Research Exceptions: The rules include exceptions for AI components provided under open source licenses and for research activities. This could stimulate innovation in these areas, but it might also create loopholes that could be exploited.

  2. AI Governance: The role of the EU AI Office in monitoring the implementation of the AI rulebook will be crucial. However, it might also be a source of potential bias, as it could favor or disfavor certain types of AI applications.

  3. AI and Democracy: The inclusion of AI systems used to influence voters in political campaigns in the high-risk category highlights the MEPs' awareness of the potential impact of AI on democratic processes. This is a highly topical issue, and its inclusion may serve as a model for other jurisdictions.

  4. The Role of Judicial Oversight: The rules allow for the use of "post" remote biometric identification systems by law enforcement for the prosecution of serious crimes, but only after judicial authorization. This highlights the important role of judicial oversight in balancing the benefits of AI with the protection of civil liberties.

  5. The Future of AI: The technology-neutral definition of AI proposed by the MEPs is designed to apply to both current and future AI systems. This indicates an awareness of the rapid pace of AI development and an attempt to future-proof the regulations.

Overall, these proposed rules represent a significant step towards addressing some of the most pressing ethical and societal concerns associated with AI.
However, their success will depend largely on the robustness of the enforcement mechanisms and the ability to balance regulation with innovation.

Our Daily Regulatory News Service

The Regulatory News Service: Grand Compliance - May 14
📌Council Decision with Union Rules on Data Protection📌 EDPS and DPO Network of EU Discuss GDPR Compliance📌 Market Vulnerabilities and Status Report📌 ICAEW AML Supervisory Activity Q1 2023📌 Rising Interest Rates and Banking Implications📌 BIS Handbook on Offline Use of CBDCs for Cyber Securit…

Get Compliant Now ⬇

Reduce your
compliance risks