Brown University Professor Discusses AI Regulation
On September 18, 2025, Professor Suresh Venkatasubramanian, an AI safety researcher at Brown University, delivered a lecture at Carnegie Mellon University focused on the crucial topic of artificial intelligence (AI) regulations and evaluating fairness in AI applications.
As the Co-Director of the newly established AI Research Institute on Interaction for AI assistants (ARIA), Venkatasubramanian brings a wealth of experience from his previous role as the Assistant Director for Science and Justice in the White House Office of Science and Technology Policy during the administration of former President Joe Biden. In this capacity, he co-authored the Blueprint for an AI Bill of Rights.
Venkatasubramanian noted a significant shift in the government’s stance towards AI regulations, transitioning from a cautious approach to a more aggressive regulatory framework. He expressed a sense of “profound depression and existential dread” regarding the future of AI governance, stemming from the rapid developments in the field and their implications.
Regulatory Landscape Changes
According to Biden’s Executive Orders in 2023, the administration imposed stricter scrutiny on AI development, particularly in terms of risk assessment and addressing structural discrimination. This marked a stark contrast to the previous administration, which had rolled back many regulatory efforts while still maintaining supportive provisions for data center infrastructure projects.
Currently, the federal government views AI as a critical aspect of global competition and has adopted a deregulatory, pro-innovation approach to expedite AI development. However, the absence of clear regulatory guidance has forced researchers like Venkatasubramanian to seek out methodologies to assess the impact of AI technologies.
Triangular Inquiry Model
Venkatasubramanian advocates for a triangular inquiry model that encompasses framing, tools, and measurement. While much of the scientific community focuses on developing technical tools, he emphasizes the need for policymakers and researchers to consider how problems are framed. He warns that failing to recognize different frames can lead to adopting perspectives that are not beneficial.
“If we are not aware of the different frames that are in place when talking about tech and tech governance, we will be trapped into a frame that someone else wants rather than a frame that we have chosen ourselves,” he remarked.
Bottom-Up Approaches
Venkatasubramanian argues that meaningful advancements in AI governance will not emerge from rigid top-down frameworks but rather from practical, bottom-up approaches that address specific issues within communities. He acknowledges that while scaling solutions is often viewed as ideal, it can sometimes hinder progress.
“I think in computer science, as a gospel, that scaling is what we want to do. We have to understand that sometimes, scaling is preventing us from doing the thing we want,” he cautioned.
Focus on Policy Targets
Regarding the formulation of AI regulations, Venkatasubramanian urges policymakers to concentrate on policy targets rather than relying on technical specifications, which can quickly become outdated in the fast-evolving landscape of AI technology.
“What you should be doing is focusing on identifying targets based on application and consequences to insulate yourself from some of the issues with generative AI that haven’t changed rapidly,” he advised.
A Critical Moment for AI Governance
Venkatasubramanian believes that now is a pivotal time to establish a foundation for effective AI governance. “We don’t have the luxury to spend three years figuring out the best vehicles. We have to work on this right now because if not, something will get set in place and will be very hard to change,” he stated.
He concluded by encouraging students and emerging professionals in the field, emphasizing that collective action is essential: “None of us have to change the world by ourselves. It’s going to be a group effort … just mind a small piece of it.”
