Experts look at the benefits and pitfalls of AI regulations.
Recently, the US House of Representatives voted in line with party lines and passed HR 1, known as “One Big Beautiful Bill Act.” If enacted, HR1 suspends AI (AI) models or conditions or local regulations affecting a decade of research.
Over the past few years, AI tools have become more popular among consumers, from chatbots like ChatGpt and DeepSeek to sophisticated video generation software like Alphabet Inc.’s VEO 3. Approximately 40% of Americans use AI tools every day. These tools continue to improve quickly and become easier to use and useful for the average consumer and business users.
Optimistic forecasts suggest that ongoing adoption of AI could lead to trillions of dollars of economic growth. However, unlocking the benefits of AI definitely requires meaningful social and economic adjustments in the face of new employment, cybersecurity and information consumption patterns. Experts estimate that a wide range of AI implementations could replace or transform approximately 40% of existing jobs. Some analysts warn that without a robust safety net or reskilling programme, this displacement could exacerbate existing inequalities, especially for low-income workers and communities of colour, and among fewer developed countries.
Given the dramatic and widespread potential for economic evacuation, domestic and state governments, human rights watchdogs and trade unions are increasingly supporting greater regulatory oversight in the emerging AI sector.
The data center infrastructure needed to support current AI tools consumes as much electricity as France’s 11th largest national market. Continuing growth in the AI ​​sector requires more slowly power generation and storage capabilities, creating great potential for environmental impact. In addition to the use of electricity, AI developments consume large amounts of water for cooling, raising concerns about further sustainability in water sculpture regions.
Industry insiders and critics likewise point out that excessively widespread training parameters and defects or representative data allow models to embed harmful stereotypes and mimic human bias. These biases lead critics to seek stricter regulations on the implementation of AI in police, national security, and other policy contexts.
Polls show that American voters want more regulation for AI companies, including limiting the training data that AI models can adopt, imposing environmental impact taxes on AI companies, and banning the implementation of AI in some sectors of the economy entirely.
Nevertheless, there is little consensus among academics, industry insiders and lawmakers about whether emerging AI sectors should be regulated.
At this Saturday seminar, academics will discuss the need for AI regulation and the advantages and disadvantages of centralized federal oversight.
In an article from Stanford University’s Emerging Technology Review 2025, Stanford University’s Fei-Fei Li, Christopher Manning, and Anka Reuel, argue that federal regulations on AI could undermine U.S. leadership by locking in strict regulations before key technologies mature. Li, Manning and Reuel have warned against centralizing regulations on general-purpose AI models, blocking competition, seizing dominant companies, and shutting down third-party researchers. Instead, Li, Manning, and Reuel are seeking flexible regulatory models based on existing sector rules and voluntary governance to address usage-specific risks. Such an approach suggested by Li, Manning, and Reuel will better maintain the benefits of regulatory flexibility, while maintaining target monitoring for the most risky areas. Philipp Hacker, a professor at the European University of Viadrina, argues that AI regulations must highlight the critical climate impacts of machine learning technology. Hackers emphasize the substantial energy and water consumption required to train large-scale generative models such as GPT-4. Hackers criticizing the current European Union regulatory framework, including general data protection regulations and the then proposed EU AI law, will encourage policy reforms that move beyond transparency to incorporate the sustainability of design and consumption caps associated with emissions trading schemes. Finally, hackers propose these sustainable AI regulatory strategies as a broader blueprint for environmentally conscious development of emerging technologies such as blockchain and metaverse. In a recent briefing paper, Insertra explains that regulatory schemes often target content labeled as misinformation or hate speech. Insertra warns that such rules could entrench dominant companies and block AI products designed to reflect a wider range of perspectives. Insertra allows for the development of AI tools that support the diverse expression of speech in pursuit of flexible approaches based on soft laws, such as voluntary codes of conduct and third-party standards. North Carolina Legal Review Article, Erwin Chemerinsky, Dean of UC BERKELEY LAGE, X-Chemiet’s Standard Standardization Constitutional Issues and Bad Policies in Standard Chemistry. Miami Herald v. Tornilho and Harley V. Utilizing precedents such as Irish-American gay groups, Kemelinsky, and Kemelinsky, claims that many state laws that limit or require content moderation are violating the protection of First Amendment Editorial Editorial discretion. Chemerinsky and Chemerinsky further argue that federal law prevails content moderation regulations in most states. Chemerinskys warns that by allowing multiple state regulatory schemes, the most restrictive states will effectively control internet speech across the country, causing the “lowest common denominator” issues that undermine platform editing rights and user free expression. Yun argues that excessive limitations in AI regulations can lead to long-term social costs outweigh the short-term benefits obtained by risking inducing innovation and reducing immediate harm. Similar to the early days of internet regulations, Yoon emphasizes that early intervention could entrench market positions, limit competition and converge potentially superior, market-driven solutions with new risks. Instead, Yun’s supporters apply existing laws of general application to AI, maintaining regulatory restrictions similar to the approach adopted in the formative early days of the Internet. A future article by Rogerscalisa at the University of Oslo will explore how AI regulations in different countries created the diversity of Uneven Storm in Uneven Storm. Kaliisa and his co-authors analyze how comprehensive EU regulations, such as AI law, US-specific approaches, and disclosure requirements for Chinese algorithms, impose various restrictions on the use of educational data in AI research. Kaliisa and his team warn that strict rules, particularly the EU’s prohibitions of emotional recognition and biometric sensors, will limit innovative AI applications and widen global inequality in educational AI development. The Kaliisa team proposes that experts engage policymakers and develop a framework that balances innovation and cross-border ethical protection measures.
Saturday seminars are weekly features aimed at writing the kind of content that will be conveyed in live seminars involving regulatory experts. Each week, the regulatory review publishes a brief overview of a selected regulatory topic and distills recent research and academic writing on that topic.
