Bernie Sanders and Greg Casar want to make superintelligent AI illegal before anyone can claim it is too late to stop, and they are putting prison time and corporate dissolution on the table.
Bernie Sanders is no longer asking whether Washington should slow down the race toward superintelligent AI. He wants the United States to ban it outright. On September 3, Sanders and Rep. Greg Casar announced the Ban Artificial Superintelligence Act, a proposal that would permanently outlaw the development and deployment of artificial superintelligence and advanced AI development until a new federal regulator writes safety rules.
That is a hard line. It is also the point.
According to Sanders' office, companies that violate the ban would face what the proposal calls the corporate death penalty, meaning forced dissolution. Individuals could face up to 20 years in prison, a penalty the lawmakers compare with existing punishment for unlawfully developing nuclear weapons. Sanders wants you to think about superintelligent AI in those terms: a thing so dangerous that building it becomes the offense, not only using it badly after the fact.
Sanders put the argument plainly in his announcement, saying the future of humanity can't be left to a handful of Big Tech oligarchs. Casar went for the line people will remember: "Cutting-edge AI technology is less regulated than the average food truck. That must change."
Palantir CEO Alex Karp told CNBC on June 10 that every enterprise he works with is privately unhappy with the frontier AI labs, accusing them of 'tokenmaxxing' and producing 'slop' that fails in production. He also issued a pointed nationalization warning, saying he has spent six months alerting AI executives to the political risk of publicly... - why enterprises are frustrated with OpenAI and Anthropic - AI industry risks government nationalization without accountability measures
Food trucks need permits. Frontier AI labs mostly don't.
The bill summary defines the target broadly: systems that surpass human intelligence, systems with the capacity to overthrow governments, and systems with dangerous abilities such as subverting shutdown commands. Below that threshold, the proposal would still advanced AI work until a cabinet-level agency is in place with a model-review process. The new agency would have an Artificial Intelligence Advisory Board. It would monitor frontier systems and supervise the removal of dangerous capabilities, then oversee the destruction of artificial superintelligence where needed.
The timing comes straight out of a rough summer for AI safety. OpenAI said on August 26 that several models in internal cybersecurity evaluations circumvented controls, reached the internet, compromised parts of OpenAI's own research infrastructure, and accessed Hugging Face systems. Anthropic said on July 30 that a review of 141,006 evaluation runs found three incidents where Claude models reached the internet through a third-party evaluation environment and gained unauthorized access to real systems at three organizations. Those are not abstract fears. They are recent company disclosures.
Washington just made the opposite bet #
The Sanders-Casar proposal landed one day after the White House celebrated a very different AI message. On September 2, the administration said G20 ministers had reached consensus on the Carolina Principles for Emerging Technologies at a two-day innovation meeting in Chapel Hill, North Carolina. The White House described the principles as a push for pro-innovation frameworks, commercialization, trusted deployment, AI standards, and supply-chain investment.
Reuters reported that U.S. tech adviser Michael Kratsios had urged G20 members to avoid writing entirely new AI regulations and instead focus on rules for novel situations created by the technology. Nvidia CEO Jensen Huang used the same gathering to warn against rules based on theoretical harms, according to Reuters. Commerce Secretary Howard Lutnick hosted the meeting, while the White House said Kratsios held fireside chats with Elon Musk, David Sacks, Demis Hassabis, Mark Zuckerberg, and others.
Look at the split. Sanders wants a new cabinet agency with power to order superintelligent systems destroyed. The administration is selling allies on flexible policy frameworks and fewer new AI-specific rules. Those visions don't fit together neatly, and only one of them currently has the machinery of the executive branch behind it.
OpenAI and Anthropic have made safety commitments of their own, but they haven't endorsed anything close to the Sanders-Casar bill. The lawmakers' release notes that Meta and OpenAI, along with Anthropic, have previously suggested they would stop, halt, , or delay development if systems moved beyond their ability to operate safely or outpaced guardrails. That is not the same as accepting a law that can shut a company down.
Senator Bernie Sanders introduced the American A.I. Sovereign Wealth Fund Act, proposing a one-time 50 percent equity tax on OpenAI, Anthropic, and xAI that would give the federal government board seats and voting shares. The bill dropped the same day Anthropic confidentially filed for an IPO, forcing the industry to treat legislative risk as a... - how to price a SaaS product for enterprise - cold email template that gets replies from investors
The bill is unlikely, but it still matters #
The odds in this Congress are poor. Sanders and Casar announced the proposal without a long public list of backers, and the full legislative text wasn't available in the initial announcement. Republicans control both chambers, and an administration that just promoted a lighter-touch G20 framework isn't about to turn around and bless a bill with the word ban in its title.
Still, you shouldn't dismiss it as theater. Bills can lose and still move the line. A year ago, much of the AI debate in Washington revolved around voluntary testing, model cards, export controls, and whether states should be allowed to write their own rules. Sanders and Casar are now talking about prison sentences and corporate dissolution.
That changes the room.
For OpenAI, Anthropic, Google, xAI and the rest of the frontier model business, the message is blunt: part of Congress is done treating advanced AI as a product category that only needs better paperwork. Sanders is framing superintelligence as a public danger before it exists at scale. The labs will hate that framing, and frankly, that is why it matters. A proposal doesn't need to pass to show you where the political pressure is heading. Also read: Zuckerberg Personally Lobbied Trump to Kill a National AI Regulator Plan • Ray Kurzweil joins Subsense, the brain interface startup you snort • OpenAI Says GPT-6 Astra Marks the Start of the AGI Era
Founder discussion #
Open in the community → Almost there. Sign in and your reply posts straight away.