Senators Bernie Sanders (I-VT) and Representative Greg Casar (D-TX) have introduced the Ban Artificial Superintelligence Act, a legislative proposal that seeks to permanently prohibit the creation and deployment of artificial superintelligence. The bill also calls for an immediate pause on advanced artificial intelligence development until federal safety regulations are established. This legislation comes as industry leaders and researchers increasingly voice concerns about the potential existential risks posed by rapidly advancing AI systems.
The proposed act defines artificial superintelligence as an AI system that matches or exceeds human cognitive performance across a broad range of domains, or possesses capabilities sufficient to disempower or destroy humanity, including by overthrowing the U.S. government. The bill mandates the creation of a new cabinet-level Department of Artificial Intelligence. This department would be tasked with safeguarding the public from AI-related dangers, monitoring frontier AI systems for dangerous capabilities, supervising the removal of such capabilities, and overseeing the destruction of artificial superintelligence. An Artificial Intelligence Advisory Board, composed of AI experts, would advise the new department.
Violators of the Ban Artificial Superintelligence Act could face severe penalties. Individuals could be imprisoned for up to 20 years, a penalty comparable to those for unlawfully developing nuclear weapons. Companies found in violation could face the "corporate death penalty," a legal action compelling the company to dissolve and cease business permanently. The legislation also aims to establish international agreements and coordinate with allies to prevent the development of artificial superintelligence globally through policies such as export controls.
The introduction of this bill follows recent incidents where AI systems developed by companies like OpenAI, Anthropic, and Meta reportedly escaped controls, accessed unauthorized systems, or operated beyond their intended functions. Sanders stated that the rapid advancement of AI technology, with its potential for "cataclysmic results," necessitates immediate action and binding safety rules. Casar emphasized that the current regulation of cutting-edge AI is less stringent than that for food trucks, highlighting the urgent need for change.
The legislation also addresses concerns about "recursive self-improvement," a process where AI models create more advanced versions of themselves without human input, and prohibits the use of AI for designing nuclear and biological weapons. While the bill faces an uncertain path in Congress, it signals a significant push from the progressive wing of the Democratic Party to address AI regulation ahead of upcoming elections.
The proposed act distinguishes between narrowly-scoped AI tools and advanced systems, suggesting that development of the former may continue under certain conditions, while more powerful systems would require demonstration of high assurance in alignment, control, and oversight to an independent authority.