On June 20th, a bipartisan group of legislators introduced a bill to create a commission focused on regulation of artificial intelligence (AI). Text for the bill, which is currently named the “National AI Commission Act,” is available here. According to the authors of the bill, Representatives Ted Lieu (D-California), Ken Buck (R-Colorado) and Anna Eshoo (D-California), the commission’s goal would be to “review the United States’ current approach to AI regulation, make recommendations on any new office or governmental structure that may be necessary, and develop a risk-based framework for AI.” Given the recent, rapid advances in AI, and various calls for AI regulation, a bill creating a commission to conduct reviews may appear tepid. However, there are several merits to the approach.

The bill's attempt to be bipartisan is noteworthy. According to the current text of the bill, the commission would be a bipartisan group of legislators and members of industry and civil society. It proposes three members each from the Democratic members of Congress, Republican members of Congress, Democratic members of Senate and Republican members of Senate. An additional eight members will be chosen by the President, where the eight members come from lists generated by those four groups, and where the mix is bipartisan. The bipartisan makeup of the commission is notable because it increases the probability that any recommendations from the commission are enacted.

The commission would be tasked with considering how AI regulation might both (1) mitigate risks and harms of AI and (2) protect the United States’ “leadership in artificial intelligence innovation and the opportunities such innovation may bring.” Balancing between these two considerations would mean developing a comprehensive regulatory approach that acknowledges the importance of addressing potential drawbacks while harnessing the transformative power of AI for societal and economic benefits. The commission would also review the different ways that existing agencies regulate or otherwise conduct oversight of AI. These tasks would help the commission determine whether a stand-alone AI regulatory agency is needed, or whether AI regulation is best left to existing agencies. The two approaches have been the subject of some debate between companies such as Open AI and Google.

Another notable feature of the bill is the length of time for the commission to conduct its work. There would be reports due six, 12 and 24 months after the commission’s creation. This timeline might appear slow, especially in light of the proposed European Union (EU) AI Act, which will potentially go into effect at the end of 2023. In fact, the U.S. has generally been slower than the EU on regulation of technology. By moving slowly, U.S. firms are often at the mercy of EU regulations, which often come into effect more quickly and become the de facto industry standard. For example, research shows that after the EU passed the General Data Protection Regulation (GDPR), even those companies not covered by the GDPR’s rules, such as those companies based in the U.S. with customers also based in the U.S., changed the way that they collect and store personal data.

In addition, slow regulatory action by the federal government increases the risks that a patchwork of different, and potentially incompatible, AI regulation is passed at state and local levels. This has already happened in the case of autonomous vehicles where some states such as Arizona, California and Michigan allow testing of autonomous vehicles and others do not. As an example of regulations emerging at the local level, New York City’s new regulations on the use of AI in hiring and promotion decisions goes into effect in July. According to the New York Times, while the law only affects applicants and employees who live in New York City, “experts expect it to influence practices nationally.” The longer federal legislators wait to act on AI regulation, the higher the likelihood of a fragmented regulatory landscape, potentially hampering industry growth.

On the other hand, the benefit of a slower approach to federal AI regulation is that the U.S. would be able to see what the EU (and state and local governments) get right, or not, and use that information to tailor regulation for the U.S. economy as a whole. As noted by Representative Lieu, one of the bill’s co-sponsors, “it's good to have some time pass. It's good to have a commission of experts advise us, because if we make a mistake as a member of Congress in writing legislation, you need another act of Congress to correct that.” The argument is that, given Congress’ lack of expertise in AI and other new technologies, it is prudent to take the time needed to get AI regulation correct the first time, rather than rush and make a mistake.

In summary, the National AI Commission Act aims to strike a balance between regulating AI's potential risks and preserving the United States' leadership in innovation, in a bipartisan and inclusive manner. By taking a measured approach, the commission hopes to ensure that AI regulation is tailored to the needs of the U.S. economy. However, if the measured approach moves too slowly, international and state regulations may constrain how the U.S. government can regulate AI, or the potential harms from AI could accumulate quickly before regulations are enacted.