The Sanders-Casar “Ban Artificial Superintelligence Act” is AI Authoritarianism
Last week, Sen. Bernie Sanders (I-Vt.) and Rep. Greg Casar (D-TX) introduced the “Ban Artificial Superintelligence Act,” a sweeping new bill that envisions comprehensive government control over AI and advanced computation. The measure contains some of the most radical interventions ever proposed in any piece of proposed legislation in American history. If enacted, it would have extremely destructive and dangerous consequences for innovation, competition, economic growth, global competitiveness, national security, and freedom of speech.
The bill’s provisions notably include up to 20-year jail sentences for certain violations, an indeterminate “pause” on AI development, and the creation of a new cabinet-level “Department of Artificial Intelligence” tasked with monitoring and controlling the technology. The bill also directs the administration to pursue international agreements with other countries regarding safety standards around AI, potentially implying restrictions on countries that continue to develop these systems outside of international frameworks.
Sanders has previously floated other proposals including the government taking a 50 percent stake in frontier AI companies, and the creation of an AI sovereign wealth fund the proceeds from which would be used to ensure that “economic benefits generated by AI are used to improve the lives of all of us.” Previous R Street essays have noted how these measures, if implemented, would not only stifle innovation and expose the federal government to potentially significant contingent liabilities, but also usher in unprecedented and dangerous control over this cutting-edge technology.
However, there exists an apparent contradiction between Sanders’ current proposal and his earlier efforts from over the summer: the former would restrict the development of AI, while the latter seek to use the economic returns generated by AI companies to fund public programs. The more aggressively policymakers constrain the development and commercialization of AI, the smaller the economic base from which such revenues could ultimately be extracted. One could be forgiven for thinking that the true purpose of these bills is not to ensure that AI is developed in a safe and equitable manner, but rather to assert government control over the technology and the firms that develop it.
Regardless, if implemented, this sort of authoritarian thinking and corresponding set of regulatory proposals would represent an unprecedented expansion of government power over technology and speech in America. While it is tempting to dismiss these ideas as too radical to be considered seriously, the reality is that many of the ideas Sanders has floated are being openly discussed now by pundits, the press, and policymakers. A new age of AI radicalism is upon us, and the dangers associated with these ideas need to be better understood.
Pause and Effect
The same day the Sanders-Casar bill dropped, a front-page story in The Wall Street Journal noted that the current projected “AI buildout” of $10.3 trillion between 2025 and 2032 is likely to be one of the largest private sector infrastructure investments since the construction of the railroads in the late-19th Century. More generally, a Federal Reserve staff report noted that investments related to this buildout have contributed meaningfully to quarterly growth in Gross Domestic Product from 2025 through the first quarter of 2026. Indeed, by one estimate, the boom has created more than 750,000 new jobs in the U.S. since 2023. Yet whether current and prospective investments can be sustained depends on future revenues, which in turn depend on the continued adoption, development, and commercialization of AI technologies and uses.
As other commenters have noted, the bill is rather vague in its definition of the systems it targets. Its central terms, including “artificial superintelligence” and “superintelligence precursor characteristics,” depend on open-ended capabilities such as the capacity to “automate or greatly accelerate” AI research. Adding to uncertainty, the bill’s compute threshold that triggers the pause can be adjusted annually by the AI Secretary. The ambiguity of a loosely-defined “pause” on AI training and modification—and conditioning the release of new models on regulatory approval—introduces just the kind of regime uncertainty that tends to stymie long-term capital investment. It also harms emergent competition from new players, especially open-source firms and platforms. Investors contemplating data centers, power infrastructure, networking equipment, and other complementary assets must not only form expectations about future demand for AI services, but also about whether the government will permit the technologies generating that demand to be developed and deployed.
Importantly, the bill does not provide any clear end date as to when such uncertainty would abate. Indeed, it only stipulates that the pause on AI development would continue until such a time as the Secretary of the proposed Department of Artificial Intelligence deems that the agency is fully operational and a new set of comprehensive AI regulations are in place. Given the current dysfunction in Congress, particularly when it comes to handling major agenda items like establishing new bureaucracies, the time that elapses between the bill’s enactment and the end of the pause could be years.
The bill’s effects would not stop at frontier labs. By pausing current AI development and subsequently limiting competition to those firms that secure a federal charter, the bill’s proposed regulatory architecture would further impede the innovation process and the discovery of new, productive uses for AI. While the European Union’s (EU) recently enacted AI framework is less restrictive than the regime proposed by Sanders-Casar, Europe is already experiencing the effects of this kind of prescriptive regime, with a recent working paper finding that the EU’s new comprehensive AI framework both reduced high-tech firms’ research expenditures and negatively affected their overall valuation, especially among smaller firms.
The effects of such regulations, however, extend beyond these first-order effects. Delayed deployment means less experimentation, less “learning-by-doing,” and fewer complementary investments in new skills, organizational capital, and infrastructure that are essential to fully capturing the benefits of novel general-purpose technologies such as AI. Compounded over time, these forgone investments could drastically attenuate the broader gains to society from technological innovation all for the sake of reducing the still uncertain risks associated with powerful AI systems. Yet, while these economic costs compound, the technological capabilities of America’s geopolitical rivals would continue to push forward.
The AI Prisoner’s Dilemma
The Sanders-Casar bill would have dangerous national security ramifications and would represent a de facto declaration of technological surrender in the AI and robotics revolution as China races ahead on both fronts. As previous R Street testimony has noted, there is a symbiotic relationship between a nation’s technology capabilities, its global competitiveness, and national security. AI is both the most important general-purpose and dual-use technology of modern times, meaning that policies that result in a diminished domestic technology base will have potentially dangerous ramifications for the geopolitical balance of power in an era of “great power competition.”
Chinese Communist Party (CCP) leaders understand this and have repeatedly articulated their intent to be the global leader in AI, robotics, and advanced computational capabilities. The CCP has simultaneously devoted considerable resources to promoting the development of its AI and robotics sector both domestically and abroad. China is already well ahead of the U.S. in the development of drones and general-purpose robotics, and is close on America’s heels in terms of frontier AI model capabilities.
If the Sanders-Casar bill were to pass, this computational capabilities race would be over, and China would easily take the global lead. The measure would directly limit the technological frontier by pausing AI development and then add layers of new regulatory requirements and bureaucratic procedures to future model innovation.
Sanders and Casar operate on the assumption that China will ultimately agree to limit its own capabilities, but there is no sign they are willing to do so. Moreover, even if CCP officials did make such promises, it is unlikely America would be able to trust them to comply, especially given their track record of not complying with nuclear non-proliferation and human rights agreements.
Finally, Sec. 15(b) of the Sanders-Casar bill measure also contains a troubling provision stipulating that, “It shall be the policy of the United States to prevent the development of artificial superintelligence globally, which may include export controls on artificial intelligence computing infrastructure.” Besides the mention of export controls, nothing more is included about how the development of artificial superintelligence will be controlled globally by Washington. Taken literally, this could be a recipe for eventual interventions and conflicts with other nations, most notably a nuclear-armed China. At a minimum, such language could be viewed as containing such a threat, thus undermining global diplomacy and peace.
Conclusion
There are still other dangers associated with the sweeping controls contained in the Sanders-Casar bill. For example, the measure would open the door to unprecedented expansion of government power over information creation and dissemination. Because the bill includes a ban on advanced AI systems, it raises classic First Amendment concerns about prior restraint and government influence over information platforms. Government control of AI development can burden the rights of individuals and organizations to build, configure, use, publish, and receive information from these systems.
Taken together, these problems with the Sanders-Casar bill and earlier proposals from Sanders point to the troubling growth of authoritarian thinking in AI policy. There are many more reasonable approaches to addressing concerns about system safety, including better enforcement of the many different laws and regulations already on the books. Some additional laws may be needed to address frontier AI safety, but not of the radical nature of the flat bans and extreme regulations contained in the Sanders-Casar bill.