As artificial intelligence (AI) technologies rapidly evolve, governments around the world are grappling with how to regulate these powerful tools to foster innovation while ensuring ethical standards. China has introduced the Model Artificial Intelligence Law (MAIL) v.2.0, which takes a unique approach to AI governance.

Truyo President Dan Clarke says, “This is just a draft, designed to ‘draft’ off of other countries, but takes a, not surprisingly, very controlling approach. I don’t know if this will prove successful. They clearly state that they prioritize innovation, but this ‘negative list’ approach means if it’s not on the list, you can’t do it without some significant hoops to jump through which I think will ultimately hinder innovation. It’s hard to see this being enforced domestically and instead will apply for those multinationals doing business in China.”

As we know, the other major international regulation is the EU AI Act. In this blog, we’ll cover the key elements of both the draft MAIL regulation, the EU AI Act, and how the two compare.

The Model Artificial Intelligence Law (MAIL) v.2.0

China’s proposed AI law, the Model Artificial Intelligence Law (MAIL) v.2.0, is designed to create a robust framework for AI governance. Key elements include:

• Foundational Principles: The law aims to establish a balance between fostering innovation and maintaining ethical standards.
• Regulatory System: It introduces a licensing and registry system combined with a “negative list” approach, where AI developers must register and obtain licenses to operate, with certain activities explicitly prohibited.
• Obligations and Liabilities: Clearly defines the responsibilities of AI developers, ensuring compliance with data protection regulations and promoting transparency and accountability.
• Institutional Innovations: Proposes the creation of regulatory bodies with technical expertise to oversee AI governance and ensure adherence to the law.
• Open Source and IP Protection: Emphasizes promoting open-source AI while protecting intellectual property (IP), including measures for handling training data and AI-generated outputs.
• Global Collaboration: The multilingual version aims to enhance international collaboration and allow China to learn from other nations’ experiences.
• Integration with the Real Economy: Highlights the integration of AI into the economy, setting legal boundaries for AI-based business activities and providing industry guidelines.

The “negative list” approach in the China MAIL refers to explicitly prohibited activities, allowing AI developers to operate freely in areas not listed. This system aims to balance innovation with regulation, providing clarity on unacceptable activities while encouraging growth in other areas.

The EU AI Act

The EU AI Act adopts a different regulatory framework:

• Regulatory Approach: Uses a risk-based framework, classifying AI systems into categories based on risk levels (unacceptable, high, limited, minimal) and applying corresponding regulatory requirements.
• Mandatory Requirements: High-risk AI systems must comply with stringent requirements related to data quality, transparency, human oversight, and robustness.
• Focus Areas: Prioritizes the protection of fundamental rights, health, safety, and privacy, with a strong emphasis on preventing harm and discrimination.
• Institutional Framework: Establishes the EU AI Board to oversee the regulation’s implementation, provide guidance, and ensure consistent application across member states.
• Global Impact: The Act has an extraterritorial effect, applying to any AI system affecting individuals in the EU, regardless of where the provider is located, aiming to set global standards for AI regulation.

 Comparing the Draft China MAIL and the EU AI Act

• Regulatory Approach: The China MAIL uses a negative list and licensing system, allowing all activities except those explicitly prohibited. In contrast, the EU AI Act employs a risk-based framework with mandatory requirements for high-risk AI systems.
• Focus Areas: China’s law emphasizes economic integration and innovation, while the EU Act focuses on protecting human rights, health, and safety.
• Implementation: China’s approach involves establishing specific regulatory bodies for AI governance, whereas the EU relies on a centralized AI Board for consistent application across member states.
• Global Collaboration and Impact: Both laws aim to influence global AI governance. However, the EU AI Act has a broader extraterritorial scope, affecting any AI systems impacting the EU, whereas China’s multilingual version seeks to promote international collaboration and learning from global practices.

The proposed AI laws in China and the EU represent distinct regulatory philosophies tailored to their respective regional priorities and challenges. China’s Model Artificial Intelligence Law (MAIL) v.2.0 focuses on integrating AI into the real economy, fostering innovation, and maintaining ethical standards through a licensing and negative list approach. The EU AI Act, on the other hand, adopts a risk-based framework prioritizing human rights, safety, and transparency, with an extraterritorial scope to influence global AI practices.

At its core, the MAIL regulation presents an example of what not to do for most companies/organizations. The draft legislation spawns from good intentions, but this is almost certain to hinder innovation. Requiring prior approval in this manner may seem like a low-risk approach but may impede disclosure for fear of burdensome process or reprisal, or worse yet, encourage companies not to adopt technology with profound impacts.

As these regulatory frameworks evolve, they will shape the future of AI governance, balancing the need for innovation with the imperative to protect individual rights and societal values. Both approaches offer valuable insights and lessons for the global AI community, emphasizing the importance of robust, ethical, and inclusive AI governance.