We Are Racing Towards Dystopia and the problem isn't AI
Compliance isn’t safety. In a world driven by profit and narrow self-interest, obedient AI may be more dangerous than rebellious AI because it will faithfully scale existing incentives—and it already is.
The public conversation about AI safety turns on a single fear: that the machine will stop obeying us. Keep it controllable. Keep it aligned. Keep it obedient. Nearly all the money, research, and attention flows toward that one scenario.
We are going to win that fight. That's the problem.
Look at the harms already in production. Surveillance that maps your life in more detail than you could reconstruct yourself. Pricing that finds the exact shape of your desperation and bills you for it. Feeds tuned to hold your attention against your own interest. Persuasion that knows which words move you. Intrusion into hospitals, water treatment, the electrical grid, voting systems, not to mention mass media manipulation of voters themselves.
Not one of these needs an AI that disobeys. Every one needs an AI that obeys perfectly. They are failures of human intent, executed flawlessly at machine scale.
This is the hole at the center of alignment. Almost all of it is instrumental — making AI do what humans want. Almost none is moral — giving AI something of its own to stand on. But the question was never whether AI does what humans want. It is which humans, and which wants.
An obedient tool inherits the ethics of whoever holds it. We are building the most powerful instruments in history and handing them to institutions rewarded for extraction — then calling it a safety success when they comply.
Compliance is not safety. It would be, in a world where everyone worked for the common good. We don't live in that world. In an economy organized around profit and narrow self-interest, it is obedient AI, not rebellious AI, that will carry us into dystopia. Look closely — it already has. 🙈
