House Representatives Jay Obernolte (R‑CA) and Lori Trahan (D‑MA) introduced a 300-page draft bill to establish a framework for the United States on Artificial Intelligence (AI). The bill comes amid an increasingly fierce debate over a range of topics, from data centers to child safety to government investment in AI companies. It also comes as a growing number of states continue to consider their own AI frameworks.
There is considerable debate over what a policy framework for AI should do. Back in 2023, I laid out four principles that I still believe would continue the US’s history of a light-touch approach to innovation policy that has allowed entrepreneurship to flourish and consumers to benefit from technology far more than a government-dictated approach. While AI technology and adoption have advanced in that time, these core principles still hold enormous relevance for today’s policy decisions.
They consist of: considering existing regulations that might already cover concerns, preventing a patchwork of state policies that could disrupt innovation, improving AI literacy and education to combat concerns about labor market disruption and misinformation, and considering safeguards for the government’s own use of data and AI to protect civil rights and civil liberties.
Many of the questions that have emerged about AI are not AI questions but rather ongoing debates about speech and youth online safety or data privacy. At an individual level, AI may change how one thinks about the risks and benefits of certain choices. On a regulatory level, the rapid advancement of AI also illustrates the pacing problem in some of these underlying areas where technology moves faster than law. As the inability to launch various AI products in Europe early on due to GDPR concerns illustrated, a regulatory approach built for one particular technology can come at the expense of innovation and better solutions in the future.
It is important that any policy framework consider not only technology as it exists in 2026 but also what a better future might be prevented by an overly prescriptive approach. Presuming to know everywhere AI is headed in 2026 is like presuming to know everywhere the internet was headed in 1996. There may be a need to address specific concerns and unique challenges, such as cybersecurity or government use, but a policy framework that presumes a general-purpose technology is free for entrepreneurial use and responds narrowly to specific targeted concerns about specific applications delivers the greatest benefits for both innovators and consumers.
The US took such an approach when the internet emerged, and that is part of what created an environment that enabled its innovators to become global leaders.
The Obernolte-Trahan bill, regardless of its specifics, is an important moment because it signals Congress recognizes that a state patchwork or a series of impermanent White House executive orders are not the right approaches to such an important issue. It is constitutionally correct that if there is to be a policy framework on AI, it is Congress’s role.
Cato’s technology policy scholars will analyze the bill’s details in the coming weeks, focusing on its potential impact on AI innovation and other principles, such as free expression.
