What are the emerging regulations to rein in AI harms?

The European Union has established its pioneering AI law, the EU AI Act, which governs the development, deployment, distribution and use of AI systems across member states. The Act takes a risk-based approach, categorizing AI systems into four risk levels: minimal, limited, high, and unacceptable. High-risk AI systems, such as those with serious implications for health, safety, or fundamental rights, must adhere to stringent requirements like activity logs, robustness, cybersecurity, and accuracy.

AI systems with unacceptable risks are banned entirely, including those that manipulate or deceive, generate non-consensual sexually explicit content, or produce child sexual abuse material.

In contrast, the United States lacks a comprehensive national AI law but has seen state-level regulations emerge. The White House is promoting light-touch, market-driven regulations through federal executive orders that may conflict with state laws. While the US does not have a single federal AI regulator, various federal authorities like the Federal Trade Commission and Department of Justice oversee relevant software and algorithmic processes.

China, lacking a single AI law, relies on a mix of regulations, binding rules, targeted measures, and non-binding technical standards overseen by the Cyberspace Administration of China. AI services that shape public opinion or mobilize society must be registered before public launch, and data used to train generative AI models must be legally obtained and non-infringing.

AI-generated content must also be clearly labeled, and prohibited content is subject to restrictions. Additionally, safeguards for AI companions, including mandatory anti-addiction breaks and age verification for minors, are in place.