US lawmakers propose sweeping AI restrictions with superintelligence ban

Legislation seeks to pause advanced AI systems until federal safety standards are established

United States Senator Bernie Sanders of Vermont and Representative Greg Casar of Texas introduced a new bill that would impose sweeping restrictions on the artificial intelligence industry, including banning the development of what is called artificial superintelligence.

The legislation, introduced on Wednesday, comes weeks after top AI industry researchers issued dire warnings that the technology could wipe out humanity within the next decade, according to a wave of industry insiders.

The nearly 20-page bill outlines an outright ban on the development of intelligence that can outperform a human being and has the power “to destroy or disempower humanity, including by overthrowing the federal government.”

The legislation would also call for a temporary pause on the development of advanced AI systems until a regulatory body that can oversee and develop safety standards is put in place, as well as establishing a cabinet-level department called the Department of Artificial Intelligence that would oversee the industry.

The proposed agency would be tasked with regulating what are called frontier models — a term referring to the most advanced AI models in development. The agency would ensure the models are not dangerous, and it would be empowered to remove the models from circulation if they do, in fact, become dangerous and supervise their destruction.

The bill would also impose steep penalties for those who violate the superintelligence ban, resulting in 20 years in prison, mirroring comparable punishments for individuals who develop nuclear weapons. It also states that individuals who violate the provisions would be barred from working in the AI industry for 10 years.

For companies, the legislation also calls for what the summary describing the legislation refers to as a “corporate death penalty,” in which charter holders could have their intellectual property and assets seized by the federal government.

The legislation calls for international coordination to institute a global ban. That would include pursuing international agreements, as well as instituting export controls on “artificial intelligence computing infrastructure”.

An uphill battle

Advocates of a pullback like the bill but are sceptical.

“[It’s] unlikely to pass but it’s courageous and much safer than the status quo,” Alex Turner, a former researcher who resigned from Google DeepMind in June, told Al Jazeera.

Concerned that “an artificial intelligence system that exhibits or can easily be modified to exhibit capabilities that match or exceed human cognitive performance and capabilities across a broad range of domains or tasks” is too broad, it might cover current systems. The second definition (disempowerment of govt) seems better,” he said.

The legislation comes as US President Donald Trump told the United Nations General Assembly on Tuesday that the US would rename artificial intelligence “super intelligence” in all US government documents.

“The United States also totally rejects any attempt to construct a globality scheme to control for the artificial intelligence being spoken of so much now, hereinafter called super intelligence. Changing the name,” Trump said in his remarks.

“The use of the word artificial makes intelligence sound fake,” Trump said.

The increased pressure on the AI industry ramped up earlier this month, when Jacob Coxon, a former AI researcher at Anthropic, announced his resignation from the AI giant in a viral social media post outlining the dangers of the technology.

“The people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt. If anything, many executives and senior researchers will couch their phrasing in the press to sound sensible – but I hear the same people express fear,” Coxon said.

The bill comes as both OpenAI and Anthropic near initial public offerings on Wall Street, poised for 2027 for OpenAI and potentially November for Anthropic.