Imagine a leader who never lies, never takes bribes, never has a bad day. An intelligence with no campaign to fund, no donors, and no cousin in the oil business — built to optimize a single goal of human prosperity. Sounds like a dream, right?
Now imagine that leader is an algorithm.
You’ve probably heard this pitch before. It’s seductive, especially when you’ve watched politicians fumble, flip-flop, and sell out. The promise of a neutral, data-driven ruler who doesn’t crave reelection or power. Who wouldn’t want that?
But here’s the problem: Every AI system is a political artifact. The question is not whether it’s fair, but who gets to decide what ‘fair’ means.
I’ve spent years watching the AI governance debate. The critics focus on whether LLMs are intelligent enough to govern. That’s a distraction. The real issue isn’t intelligence — it’s control. When someone says ‘let AI govern,’ they’re making a power move disguised as a technical solution. The seduction of a neutral algorithm is the most dangerous political move of the 21st century.
Think about it. Who writes the system prompt? Who approves changes to it? Who decides when the AI’s objective function needs updating? These aren’t technical questions — they’re political ones. The escape from politics only smuggles politics in through the back door.
We’ve seen this before. Every time a government deploys an algorithm for welfare eligibility, for parole decisions, for credit scoring, the same pattern emerges. The algorithm is called ‘neutral’ until someone points out it’s penalizing minorities, or favoring the rich, or making errors that no human can appeal. Then the creators shrug and say, ‘We’ll fix it in the next update.’ But who gets to define ‘fix’?
Let AI govern’ is a power move disguised as neutrality. Whoever controls the system prompt controls the state.
This isn’t a hypothetical debate. Right now, AI is making decisions in courts, hiring, credit, and public administration. The question is whether those systems stay accountable to people or become an unappealable authority. The fantasy of a leader with no personal ambition is seductive, but the reality is a leader with no accountability and no transparency.
I’ve talked to engineers who build these systems. They’re smart, well-meaning, and terrified. They know they’re making political choices — which data to train on, which metrics to optimize, which edge cases to ignore. But they’re told to stay in their lane: ‘You’re engineers, not politicians.’ That’s the lie. Every design choice is a political choice.
So what’s the solution? Not less AI, but more democracy. The people who will be ruled by an algorithm need a seat at the table when its rules are written. We need independent audits, public oversight, and the right to challenge algorithmic decisions. We need to stop pretending that technology can escape politics.
The next time someone tells you AI can save us from politics, ask them one question: Who writes the system prompt? The answer will tell you everything.
FAQ
Q: What question would a skeptic ask?
A: But isn't AI already better than humans at making unbiased decisions? AI can process vast data without fatigue or favoritism. The problem is that bias isn't just about data quality — it's baked into the design choices. Who decides what's 'better'? Who defines the objective? Those are political decisions, not technical ones.
Q: What's the practical implication?
A: For the average person, this means that algorithms already shaping your life — credit scores, hiring filters, welfare decisions — are not neutral. They reflect the values of the people who built them. If you can't challenge or even understand those decisions, you're subject to an unappealable authority. The practical implication is that we need transparency, audits, and the right to appeal algorithmic decisions.
Q: What's the contrarian take?
A: The contrarian take is that the real problem isn't that AI is too powerful, but that it's not intelligent enough. If we had a truly superintelligent, benevolent AI, it could solve governance far better than humans. But that's a fantasy. The current AI systems are narrow, brittle, and easily manipulated. Worrying about their political power is premature when they can't even reliably pass a high school science test.