Sanders’ Proposed AI Ban Faces Uphill Battle in Congress
Senator Bernie Sanders’ ambitious proposal to ban advanced artificial intelligence systems, introduced in September, is encountering significant legislative hurdles. Even with Democrats securing a majority in both the House and Senate following recent elections, the path forward for the bill appears challenging. In the Senate, the legislation would require 60 votes to overcome a filibuster, and with a narrow majority, this threshold seems difficult to reach. Furthermore, former President Donald Trump holds veto power, which could block the bill if it were to pass Congress. Consequently, analysts suggest that Democrats are more likely to pursue hearings and investigations into AI governance in the interim.
Legislative Landscape and Key Obstacles
The “Banning Advanced Artificial Intelligence Act,” co-sponsored by Senator Sanders and Representative Jake Auchincloss (not Casar as previously stated in some reports), was introduced in September. While NBC’s election results indicate a Democratic majority in both chambers, House Democratic leadership, including Minority Leader Hakeem Jeffries, has yet to publicly endorse the proposed ban.
Even if the bill were to pass the House, the Senate faces a 60-vote requirement, and the potential for a presidential veto looms large, making the bill’s passage far from certain. The House version of the bill currently has 12 co-sponsors, all Democrats, while the Senate version has yet to attract any co-sponsors and remains in committee.
Sanders’ Post-Election Agenda and Focus on Data Centers
Senator Sanders outlined a five-point agenda for Congress following the election, focusing on universal healthcare, a $20 minimum wage, overturning the Citizens United Supreme Court decision to end billionaire-funded Super PACs, ending the war with Iran, and halting military aid to the Israeli government. Notably, the AI ban was not explicitly mentioned in this post-election agenda.
More recently, Sanders has shifted his public focus to data centers. In a November 4th post, he highlighted that 43 out of 44 communities in local votes had called for a moratorium on data center construction, advocating for an “immediate nationwide moratorium on data centers.”
Defining “Advanced AI” and Potential Impact
The proposed legislation defines “advanced AI systems” as those requiring more than 10^25 floating-point operations (FLOPs) for training, a metric measuring computational effort. According to an analysis of public data from Epoch AI, at least 27 models meet this computational threshold, with projections indicating a significant increase in the coming years: 4 in 2023, 6 in 2024, 9 in 2025, and 8 so far in 2026. The actual number is likely higher, as many cutting-edge models do not publicly disclose their computational power.
If enacted, the ban would mandate a halt to the training, modification, or fine-tuning of advanced AI systems, with exceptions only for efforts to remove dangerous features or disable the systems. New, unreleased advanced models would be prohibited from going online. While existing deployed models would not be required to be taken down, they could not be further modified. The moratorium would remain in effect until a newly established AI department is staffed and its regulations are finalized. This would effectively pause the training and release of new models by major AI companies.
Beyond a Moratorium: The “Superintelligence” Ban
In addition to the moratorium, the bill proposes a ban on the development, deployment, possession, funding, or transfer of “superintelligence.” This also extends to AI exhibiting any “precursor characteristics,” such as significantly accelerating AI research, attempting to circumvent human shutdown commands, assisting in the creation of nuclear, biological, or chemical weapons, or deceiving and evading human oversight. Any AI found to exhibit such traits would be immediately disconnected from the internet. Developers would have 30 days to remove the problematic feature; otherwise, the system would be deactivated.
Companies found in violation would face severe penalties, including the revocation of their operating licenses. Their intellectual property and assets would be transferred to the federal government, with the stated aim of ensuring the complete destruction of the systems and hardware. Proponents have described this as an “corporate death penalty.”
