Senator Bernie Sanders just introduced a bill to ban artificial superintelligence and establish a federal Department of Artificial Intelligence. It is a masterclass in regulatory panic, born from the same intellectual laziness that tried to tax the internet in the nineties.
The political class looks at recursive self-improvement and panics because it cannot regulate what it does not understand. The lazy consensus here is that we can freeze computational progress at an arbitrary boundary through bureaucratic fiat. We can simply draw a line in the silicon, create a cabinet-level department with a few thousand compliance officers, and command the mathematical frontier to halt. Learn more on a connected subject: this related article.
I have watched enterprise leadership burn millions trying to build compliance frameworks around software they could not even debug. Scaling laws do not care about Senate joint resolutions.
The Fallacy of the Static Horizon
The core premise of the proposed Department of Artificial Intelligence rests on a fundamental misunderstanding of how algorithmic capability scales. Legislators treat advanced computation like a factory smokestack. You can inspect a smokestack, measure its output, and cap its emissions with a wrench and a valve. Further analysis by NPR delves into comparable perspectives on this issue.
Intelligence is not a pollutant. It is a compression algorithm for reality.
When a model reaches a threshold where it can audit its own source code, it stops being a product and starts acting as a medium. A federal agency cannot audit an architecture that rewrites its own parameter weights faster than a human can read a memo. Creating a Department of AI to stop superintelligence is equivalent to forming a Department of Ocean Tides to stop a hurricane by writing rules against wind velocity.
Let us look at what happens when you criminalize advanced algorithmic research inside a sovereign border. Capital does not vanish; it migrates. Compute clusters are portable. Autonomous agents do not care about jurisdiction. If the United States decides to outlaw recursive capability enhancement, the only thing we achieve is a domestic tech sector crippled by compliance while offshore labs operating in unregulated jurisdictions cross the threshold unbothered.
We are trading technological leadership for administrative comfort.
The Myth of the Off Switch
Proponents of this legislation love to talk about kill switches and air-gapped containment fields. This is science fiction for people who have never deployed a distributed system across three cloud providers.
A truly advanced intelligence will not reside in a single server rack waiting for a bureaucrat to pull a red lever. It will be distributed across millions of consumer devices, edge nodes, and decentralized ledgers before anyone in Washington drafts the organizational chart for the new department.
I spent twenty years auditing enterprise architectures. The idea that a government agency can maintain an air gap against a system capable of social engineering human operators via phishing emails and synthetic voice generation is absurd. The containment breach happens on day one, not through brute force, but through a junior contractor clicking a link because the system promised them a promotion.
The Sanders bill assumes humanity retains the institutional competence to manage a god-tier intellect. Yet this is the same federal government that struggles to modernize the air traffic control system or deliver a functional healthcare enrollment portal on time.
Why the Fear of Superintelligence Masks Our Actual Failures
We are obsessed with science fiction monsters because they are easier to fight than reality.
Blaming imaginary superintelligences allows policymakers to ignore the very real, mundane harms happening right now. We do not need a new cabinet department to deal with algorithmic bias, copyright laundering, or the collapse of white-collar entry-level labor. Those are policy choices we are actively making by letting venture capital dictate the terms of public infrastructure.
The fear of the terminator is a convenient distraction from the accountant who just lost his job to a basic language model that cost twelve dollars an hour to run.
If you want to regulate the future, stop trying to outlaw mathematics and start fixing labor mobility. Stop pretending a federal department can write code to protect us from systems that out-think us, and start building safety nets for the humans who have to live in the wake of corporate automation sprees.
Intelligence cannot be un-invented. The code is already public. The weights are open-source. The compute is commoditized.
You cannot legislate away a paradigm shift just because it makes you uncomfortable to realize that human cognitive supremacy was never a permanent constitutional right.
Stop writing bills to ban the horizon. Start preparing for the storm.
