US States Take a Page from the EU’s AI Act
As the United States grapples with the regulation of artificial intelligence (AI), a divide has emerged between the federal and state governments. Some commentators suggest that the EU AI Act is being replicated across the Atlantic, although there are critical differences in state legislative efforts regarding biometrics.
Federal vs. State Approaches
At the federal level, the primary concern is to avoid constraining the economic potential of AI. In contrast, U.S. states are adopting a more pragmatic approach, focusing on the regulation of AI technologies that could pose risks to society.
Colorado has taken the lead by passing SB 205 last year, although the Act remains unimplemented due to uncertainties surrounding its execution. Other states, including California, Connecticut, Iowa, Illinois, Maryland, Massachusetts, Nebraska, New Mexico, New York, Oklahoma, Texas, and Virginia, are also considering similar proposals.
Common Ground in Legislative Efforts
These legislative efforts share a common objective: to limit the potential harms of AI, particularly concerning algorithmic discrimination. Such harms are primarily identified through impact assessments, which evaluate the risks associated with AI technologies.
Concerns have been raised about the influence of the EU on American policy, referred to as the Brussels Effect. Critics argue that many algorithmic discrimination bills are adaptations of significant components of the EU AI Act, which may result in high compliance costs and increased litigation for American businesses.
Limited Implications for Facial Recognition
One of the most notable outcomes of the EU AI Act for biometric technology suppliers is the prohibition on the use of real-time remote biometric identification in public spaces. However, state-level proposals in the U.S. do not include such measures, focusing instead on ensuring compliance with existing laws before deployment.
Facial recognition developers are proactively addressing demographic differentials early in the process by tackling imbalances in datasets used for training their models. Despite concerns surrounding the use of facial recognition technology, statistical comparisons between wrongful arrests involving facial recognition and those that do not have not been substantiated, suggesting that the issue may not be as severe as some have indicated.
Conclusion
In summary, while U.S. states are drawing inspiration from the EU AI Act, they are adopting a more nuanced approach to regulating biometrics and AI. The focus remains on preemptive compliance and addressing algorithmic discrimination, but the lack of stringent measures regarding real-time biometric identification indicates a more lenient regulatory landscape than that in the EU.
