The U.S. Democratic Party recently unveiled a highly alarming draft bill—pushing the "Ban on Artificial Superintelligence Act" to completely ban the development and deployment of super AI. The penalties are striking: companies involved would be forcibly dissolved outright, and individual violators could face up to 20 years in federal prison, with sentencing standards explicitly benchmarked against serious crimes such as "illegally developing nuclear weapons." In other words, in the eyes of the proposers, secretly creating an intelligence that surpasses humans is almost the same crime as secretly building nuclear weapons. However, the bill is currently still a draft and remains some distance from actually becoming law.
What Is "Superintelligence"? For the banned artificial superintelligence (ASI), the bill provides two clear definitions. First, an AI architecture that meets or exceeds human cognitive performance across a wide range of domains and tasks, or a system that could demonstrate such capability with only slight modification; second, a system capable of planning and carrying out attacks sufficient to weaken human dominance, or even overthrow or destroy the U.S. government system. On this basis, the draft calls for a complete halt to all advanced AI research and development until a dedicated federal-level regulatory agency is in place and clear regulations and model review mechanisms are established. This amounts to pressing the pause button on frontier AI R&D until a regulatory framework takes shape.
A Cabinet Department and a Global Ban To ensure the ban takes effect, the bill proposes creating a new cabinet-level Department of Artificial Intelligence, tasked with regulating the entire life cycle of frontier models, with the power to force technical teams to remove high-risk capabilities from systems—including ignoring human shutdown commands, launching unauthorized cyberattacks, and loss-of-control behaviors such as recursive self-improvement. More notably, the ban extends its reach abroad: the draft lists "a global ban on superintelligence" as a U.S. foreign policy goal, requiring the government to seek international agreements, enlist allied cooperation, and impose strict export controls, with the aim of cutting off the possibility of ASI development anywhere in the world.
