The debate over artificial intelligence in the United States has entered a more consequential phase.
With Senator Bernie Sanders and Representative Greg Casar introducing legislation that would permanently ban artificial superintelligence, temporarily pause advanced AI development and establish a new cabinet-level Department of Artificial Intelligence.
The proposal, introduced on September 23, 2026, reflects growing concern in Washington about whether increasingly capable AI systems can remain under meaningful human control. At the center of the bill is a distinction between conventional AI and what the legislation calls artificial superintelligence.
Register for the next Tekedia Mini-MBA.
Register for Tekedia AI in Business Masterclass.
Join Tekedia Capital Syndicate and co-invest in great global startups.
The proposal would prohibit systems that exceed human cognitive performance across most domains or possess capabilities that could destroy or disempower humanity, including potentially overthrowing a government.
It would also restrict dangerous capabilities such as systems capable of circumventing shutdown commands or conducting unauthorized cyberattacks. The legislation goes beyond an outright prohibition.
Sanders and Casar are proposing a pause on advanced AI development until a federal regulatory structure establishes testing, safety requirements and model-review procedures.
The proposed Department of Artificial Intelligence would become responsible for monitoring frontier systems throughout their lifecycle and supervising the removal of dangerous capabilities. Under the bill, the department could also oversee the destruction of prohibited artificial superintelligence systems.
That approach represents a significant expansion of federal involvement in AI. Rather than treating artificial intelligence primarily as a technology-sector issue, the proposal would place frontier AI within a dedicated cabinet-level institution.
An advisory board of AI experts would provide scientific and technical guidance, while the department would have enforcement responsibilities. The bill would also establish severe penalties for violations, including up to 20 years in prison for individuals and a proposed corporate death penalty for entities that deliberately circumvent the restrictions.
The proposal arrives as AI companies are simultaneously investing enormous resources into increasingly capable models.
Sanders has argued that voluntary commitments from technology companies are insufficient, pointing to warnings from researchers and industry figures about the possibility that AI capabilities could advance faster than safety mechanisms.
His position is that government must establish binding rules before systems become too difficult to control. Yet the legislation also raises fundamental questions about innovation, competition and enforcement.
Defining the precise boundary between advanced AI and superintelligence could become one of the most difficult technical and legal challenges. AI capabilities do not develop along a single measurable dimension, and determining when a model crosses a statutory threshold could require continuous scientific assessment.
There is also an international dimension. The bill calls for the United States to pursue agreements with other countries, coordinate with allies and use measures such as export controls to prevent superintelligence from being developed elsewhere.
That reflects the reality that AI development is global: a prohibition limited to American companies would not necessarily prevent researchers or companies in other jurisdictions from pursuing similar technologies.
The proposal therefore places a much broader question before policymakers: whether advanced AI should primarily be governed as an engine of technological and economic competition or as infrastructure carrying potentially systemic risks.
The Sanders-Casar bill represents one legislative answer, but it remains a proposal rather than existing U.S. law. As artificial intelligence becomes increasingly embedded in finance, cybersecurity, scientific research, manufacturing and government.
The debate is moving beyond whether AI should be regulated. The emerging question is how much control society should retain over systems that may eventually possess capabilities exceeding those of their human creators.