The New Era of AI Regulation: Exploring the EU’s AI Act and Biden’s Executive Order

The European Union (EU) has made significant strides in the regulation of Artificial Intelligence (AI) with the recent updates to the EU AI Act. This act represents a pioneering effort to create a legal framework for the development and use of AI technologies, addressing a range of concerns from data privacy to ethical implications.

After having covered the EU AI Act in one of our recent articles, we now have some updates, which we will address in the following. Furthermore, we will compare the EU approach to the US and give an outlook on the implications current AI regulation will have on the industry.

Source: https://www.euractiv.com/section/politics/news/eu-commission-expects-first-bloc-wide-ai-law-to-be-adopted-this-year/

Key Aspects of the Act

The Act introduces a nuanced, risk-based categorization for AI systems, dividing them into four levels based on the potential risk they pose.

• Unacceptable Risk: AI systems posing clear threats to safety or fundamental rights are banned.
• High Risk: AI in critical areas like healthcare, policing, or transport requires strict compliance with regulatory standards.
• Limited Risk: AI applications like chatbots must be transparent about their AI-driven nature.
• Minimal/No Risk: AI in non-critical areas faces minimal regulatory oversight.

In terms of data governance, the Act emphasizes the need for transparency and accountability. It mandates clear disclosure when AI is involved in content generation and requires detailed documentation of the data used in training AI systems. This measure aims to address concerns around copyright and rights management.

A significant aspect of the Act is its stance on biometric surveillance and social scoring. It restricts indiscriminate biometric surveillance and outright bans social scoring practices by governments or private entities, citing privacy and ethical concerns. This move is seen as a critical step in protecting individual rights and freedoms in the digital age.

To ensure compliance, the Act establishes a dedicated EU authority responsible for overseeing its implementation and adherence, ensuring uniform application across member states.

Industry Implications

The EU AI Act, with its comprehensive regulatory framework for Artificial Intelligence, presents a range of implications for the AI industry. These implications are multifaceted, affecting various aspects of AI development, deployment, and management.

Strategic Shifts and Compliance Challenges

• Businesses that have heavily invested in technologies now classified under prohibited categories, such as biometric categorization and emotion recognition, face a critical juncture. The Act’s restrictions on these technologies necessitate major strategic shifts. Companies must reevaluate their technology focus and possibly pivot away from certain AI applications, aligning their strategies with the new regulatory landscape.

Transparency vs. Intellectual Property

• The Act’s enhanced transparency requirements pose a unique challenge, particularly in balancing the need for disclosure with the protection of intellectual property. Companies are required to be more open about how their AI systems operate, which could potentially expose sensitive trade secrets. This requirement demands a delicate balance between complying with regulatory mandates and safeguarding proprietary technologies and methodologies.

Investment in Data Quality and Bias Management

• The Act also underscores the importance of high-quality data and advanced bias management tools. Companies may need to invest more in these areas, potentially increasing operational costs. However, this investment is not without its benefits, as it enhances the fairness and quality of AI systems, aligning them with ethical standards and reducing the risk of harmful biases.

Administrative Burden and Market Dynamics

• Documentation and record-keeping requirements, as stipulated by the Act, will impose a significant administrative burden on companies. This could affect the time to market for new AI products, as companies navigate the additional layers of compliance and documentation. The need for meticulous record-keeping and reporting may slow down the innovation cycle, impacting the speed at which new AI technologies are introduced to the market.

Human Oversight in System Design

• Integrating human oversight into high-risk AI systems, as required by the Act, necessitates changes in system design and deployment. This integration may also require staff training to ensure that human overseers are adequately equipped to manage and control AI systems. This aspect of the Act emphasizes the human-in-the-loop approach, ensuring that AI systems remain under human control and oversight, especially in critical applications.

Financial Risks from Non-Compliance

• The Act introduces substantial fines for non-compliance, representing a significant financial risk for companies. These fines are intended to ensure adherence to the Act’s provisions, incentivizing companies to comply with the regulatory framework. The financial implications of non-compliance underscore the importance of thorough understanding and adherence to the Act’s requirements.

Legal Advisory and Preparedness

• Given the complexities and potential impacts of the Act, it is crucial for companies to seek legal advice and prepare adequately. As the regulatory landscape evolves, staying informed and compliant becomes essential. Companies must proactively engage with legal experts to navigate the Act’s provisions, ensuring that their AI applications and strategies are in full compliance.

AI regulations in the United States

Besides the EU being the front runner with regard to AI regulation, the US, under President Joe Biden’s administration, also initiated steps towards regulating Artificial Intelligence (AI) through an executive order — matching the EU’s efforts in addressing the challenges and opportunities presented by AI technologies.

Lets have a look at the key aspects of the executive order:

1. Government-Centric Focus: The US executive order primarily targets the deployment of AI within government sectors, emphasizing security as a core consideration. This approach marks an initial step in setting a broader policy agenda for AI governance in the US.
2. Sectoral Approach to AI Governance: The order mandates every government agency to examine AI’s relevance to their policy and regulatory jurisdictions, advancing a sectoral approach to AI governance. This includes considerations of data privacy and a call for Congress to pass relevant legislation.
3. International Engagement and AI Ethics: The executive order highlights the importance of international engagement and establishing AI ethics, aligning with global trends and geopolitical tensions. It emphasizes the US’s intent to lead the global conversation on AI ethics by example.
4. Regulatory Burdens and Challenges: The executive order introduces regulatory burdens on AI, emphasizing safety, privacy, equity, and consumer protection, essential for building trust in AI technologies.
5. Comparison with EU’s AI Act: Unlike the EU AI Act, which is legislation with enforcement, the US executive order relies on the market influence of the federal government. This difference highlights a more market-driven approach in the US compared to the EU’s regulatory-focused strategy.

Comparing the EU Act to the US AI Regulation

The EU AI Act and the US executive order on AI represent two different approaches to AI governance. With the AI Act, the EU’s approach is more regulatory and comprehensive, categorizing AI systems based on risk and imposing strict compliance requirements, especially for high-risk AI applications. In contrast, the US approach, as outlined in the executive order, focuses more on setting a policy agenda and guiding principles, particularly for AI deployment within the government sector.

Both the EU and the US recognize the importance of international engagement and ethics in AI, but their methods of implementation and enforcement differ. The EU’s approach is more prescriptive and enforceable, while the US leans towards a sectoral and principle-based approach.

Eventually, any regulation will be a knife-edge ride between responsible development, use of the technology and maintaining the competitiveness of companies in the respective jurisdictions. However, at the end of the day, we believe that a well-developed regulatory framework will provide benefits for companies due to legal certainty — hence we look forward to the developments to come.

About the author:

Guido Perscheid works as a Business Development Manager for the Center for Deep Tech Innovation and is pursuing his PhD at the Chair of Business Informatics at the University of Bamberg. Guido is an expert in innovation management with a focus on technology implementation in the fields of Blockchain, Artificial Intelligence, Internet of Things and Climate Technology.