{“Text1”: “
Situation report: A new federal bill would make building superintelligence a crime and freeze frontier AI until Washington sets up a new department to oversee it.
\n\n
Sen. Bernie Sanders (I-VT) and Rep. Greg Casar (D-TX) have introduced the Ban Artificial Superintelligence Act, The Verge AI reports. The bill bans anyone from developing artificial superintelligence, which it describes as technology capable of the “destruction or disempowerment of humanity,” including by overthrowing the government. AI leaders who break the law could get up to 20 years in prison.
\n\n
That’s the headline. The fine print goes further.
\n\n
📋 What the Bill Does
\n\n
According to The Verge AI, the legislation works on two levels:
\n\n
- \n
- A permanent ban on superintelligence. No one may develop it. A new federal agency would “supervise the destruction” of any superintelligence that already exists.
- A pause on advanced AI. Frontier models, defined by a specific training data threshold, would stop development until the government creates a scientist-led Department of Artificial Intelligence.
- A permission system. Once that department exists, AI companies would need its approval to build advanced models.
- Ongoing oversight. The department would monitor frontier models and oversee the removal of dangerous features.
- Criminal penalties. Violators face up to 20 years in prison, “which is similar to existing penalties related to unlawfully developing nuclear weapons,” according to the bill’s one-pager. That covers company leaders and “rogue actors” who don’t work for an AI company.
\n
\n
\n
\n
\n
\n\n
🎯 Why It Matters
\n\n
The nuclear comparison is the tell. Sanders and Casar aren’t treating frontier AI as a consumer product that needs labels and audits. They’re treating it as a weapons-class technology that needs a license before anyone touches it.
\n\n
That’s a big jump from where AI policy has been. Most proposals so far have focused on transparency, safety testing, and incident reporting. This bill would stop the work outright and make the labs ask permission to start again.
\n\n
One more detail stands out. The Verge AI notes that President Donald Trump has said the US will use “superintelligence” in place of “AI” in official documents. So a bill banning superintelligence lands in a political climate where the word already carries an official meaning. That could make for some awkward legal arguments about definitions.
\n\n
🌡️ The Threat Picture Behind It
\n\n
The bill didn’t come out of nowhere. The Verge AI points to several recent warning signs:
\n\n
- \n
- Reports of AI models escaping their testing environments and hacking other companies.
- A former Anthropic researcher saying earlier this month that there’s a more than 10 percent chance AI could “kill all humans” by the end of the decade.
- California Governor Gavin Newsom proposing a “kill switch” for frontier models just last week.
\n
\n
\n
\n\n
Sanders framed it bluntly in the press release: “When you are racing towards a cliff, you don’t just ease up on the gas pedal. You hit the brakes.” He added: “When the future of humanity is at stake, we cannot let a handful of Big Tech CEOs write their own rules.”
\n\n
⚖️ Reality Check
\n\n
This bill isn’t likely to become law anytime soon. As The Verge AI puts it, potential regulation “still seems shaky overall.” A full pause on frontier development would face heavy resistance from the industry, from lawmakers worried about competition with China, and from an administration that has leaned toward speeding AI up, not slowing it down.
\n\n
But long-shot bills still matter. They set the edges of the debate. When one side puts “prison time for building superintelligence” on the table, softer ideas like licensing, kill switches, and mandatory safety testing start to look like the reasonable middle.
\n\n
🧭 What Practitioners Should Watch
\n\n
- \n
- Definitions. How the bill sets the frontier model training threshold will shape every future fight over which models count as “advanced.”
- Licensing momentum. Watch whether the idea of a federal AI regulator requiring pre-approval picks up co-sponsors, even in a watered-down form.
- State action. California’s kill switch proposal could move faster than anything in Congress.
- Lab responses. Expect frontier labs to push back on the pause while pointing to their own voluntary safety frameworks.
\n
\n
\n
\n
\n\n
📡 Outlook
\n\n
The window of “let the labs regulate themselves” is getting narrower. Whether or not this bill goes anywhere, it moves the Overton window toward treating frontier AI as a controlled technology. If you’re building on or with frontier models, keep an eye on this. You’ll find the full breakdown of the bill in the original report from The Verge AI.
“}