DEMOCRACY IN THE ERA OF SUPERINTELLIGENCE
PART ONE: MESSY DEMOCRACY
The Mess We Should Protect
Perhaps the goal should never be to make democracy perfectly tidy.
A perfectly tidy political system should probably make us suspicious.
Free societies are noisy because free people disagree. The better question is what kind of mess we are looking at.
There is the mess created by argument, compromise, competing interests, investigative journalism, independent courts, elections and citizens insisting upon being heard.
That mess should not frighten us.
Then there is the mess created by corruption, deception, intimidation, contempt for law and the gradual weakening of restraints on power.
That should concern us.
But another kind of challenge may be approaching, different from either of these.
For most of democratic history, government operated at a roughly human speed. We read, argued, organized and deliberated. We wrestled with difficult choices, sometimes changed our minds—and eventually, we decided. Our institutions developed around those limitations.
AI is beginning to change all of that. We can already see how quickly these systems handle enormous amounts of information. As they improve, governments may increasingly turn to them for problems that would take human beings far longer to work through, or that may simply be too complicated for any one committee to fully grasp.
Yielding to that temptation would be completely, painfully understandable
Why tolerate months of parliamentary argument if a machine can model the consequences in seconds? Why endure committee hearings, political bargaining and public disagreement if an extraordinarily capable system can identify what appears to be the optimal answer?
There is something seductive about that possibility.
But public policy has never been merely a search for the most efficient solution.
How much privacy should we surrender for security? How should prosperity be distributed? What obligations do we truly owe to generations not yet born? And when should individual liberty yield to the greater good of the community?
Intelligence can certainly illuminate the facts behind those questions. It can spot consequences human beings might easily miss and expose choices we hadn't even considered.
But it cannot legitimately decide what we ought to value.
That distinction may only grow harder and harder to preserve.
The danger need not begin with a machine seizing political power. It could begin much more quietly, with humans gradually deciding that the machine simply makes better choices than we do.
One recommendation becomes another. Deference becomes habit. Elected representatives might still hold the formal authority to decide, while finding themselves increasingly reluctant to question conclusions they could neither independently reach nor fully evaluate.
This is where messy democracy collides with the edge of the machine.
The democracy of the future may face a challenge our constitutions were never written to handle: not simply how to restrain human power, but how to preserve meaningful human agency when intelligence itself is no longer exclusively human.
That question takes us beyond the democracy we know.
And toward the democracy we may someday have to invent.
PART TWO: THE EDGE OF THE MACHINE
Democracy in the Era of Superintelligence
For the entirety of democratic history, we’ve operated on an assumption so obvious we never had to name it—that the participants are human.
Human beings governed, advised those in power and tried to persuade one another. However sophisticated our institutions became, the decisions ultimately rested with people who shared much the same human limitations. There was only so much information we could absorb, remember and consider before making a decision.
Our democratic institutions were built around those very limitations. Legislatures deliberated. Committees investigated. Experts advised. Journalists questioned. Citizens organized. Elections periodically settled, at least for a time, the argument over who should govern.
The process moved at human speed because, in the end, so did we.
Artificial intelligence is beginning to change that familiar equation. Even today, machines can process more information than any one person could reasonably absorb. If tomorrow's systems evolve from assistants into architects of analysis, prediction, and strategy that outstrip our own, democracy will enter uncharted territory—forced to answer a question it has never faced before.
We have no idea if or when superintelligence will actually arrive. Given that uncertainty, we should be very careful about trying to predict what it will do.
But the political question is worth considering before the technology forces us to confront it:
What happens to self-government when the most capable intelligence in the room is no longer human?
The Temptation of Algorithmic Technocracy
The first pressure may not look anything like tyranny. It may simply look like better government.
Think of the many things a government has to weigh at the same time—taxes, housing, transportation, energy, public health and the economy. Now imagine an AI trying thousands of different approaches and showing us consequences that a committee of people might never have thought of.
Suppose its recommendations repeatedly produced better results.
Initially, there would seem to be little difficulty. The machine would advise and human beings would decide. We already rely upon experts, economic models and computers to help governments make decisions.
But after years of seeing the system get difficult questions right, our relationship with it might begin to change.
After all, why push back against guidance that has proven time and again to outstrip our best efforts?
This is where technological capability begins to encounter democratic legitimacy.
An advanced system might determine which policy would most efficiently increase national income. It still could not decide whether maximizing national income should be society's overriding objective.
It might calculate how much surveillance would most effectively reduce crime. Even so, the calculation couldn't tell us how much privacy a free society ought to surrender in exchange for greater security.
A system could easily calculate a healthcare model that achieves maximum measurable benefit. Even then, we would still have to decide what is fair, especially when the most efficient answer may not be the most compassionate one.
These are not failures of intelligence. They are questions of an altogether different order..
Politics, after all, is full of disagreement. We argue about the facts and about what might work. But sooner or later, we also have to decide what really matters to us and what kind of society we want.
And I think that difference will become harder to ignore as these machines grow more capable.
When people talk about the dangers of advanced AI, the conversation often turns to machines taking control from us. But that may not be the only possibility worth considering.
The transfer could be gradual and entirely voluntary.
One recommendation proves better than ours. Then another. Dependence grows because it works. Eventually we may still possess the formal authority to make decisions while becoming increasingly reluctant to exercise judgment contrary to the intelligence advising us.
No coup would be necessary.
Legislatures could remain. Elections could continue. Human beings could still occupy the offices of government.
And yet the meaning of self-government could quietly have changed.
Perhaps the difficult question is not whether machines will someday take political power from us.
It is whether, persuaded by their superior intelligence, we might gradually begin giving some of it away.
POSTSCRIPT — October 2, 2026
There is an irony I could not have anticipated when I began this reflection.
On September 29, the White House issued an executive order titled Inaugurating the Era of Super Intelligence. It directs executive departments and agencies to use “Super Intelligence” and “SI” in place of “Artificial Intelligence” and “AI” in official correspondence, public communications, websites, reports, policy documents and other non-statutory materials.
The terminology has changed. The underlying legal definition, at least for now, has not. The order applies the new name to technologies already covered by the existing federal statutory definition of artificial intelligence and directs the President's science and technology adviser to propose a federal definition of “Super Intelligence” within sixty days.
That distinction is worth noticing.
For years, superintelligence has generally pointed toward something still over the horizon: machine intelligence surpassing human intelligence broadly enough to alter our understanding of the relationship between human beings and machines.
Now a government has begun using the language before it is clear that the technological threshold traditionally associated with that language has been crossed.
This does not mean superintelligence has arrived.
Still, it makes the question I have been asking about democracy feel a little less distant.
Calling these machines superintelligent changes the way we look at them. It can also change how much we are prepared to trust them. So before that language becomes commonplace, I think we need to know what these systems can really do. Just as important, we need to know where we still want human judgment to have the final say.
The machine may not yet have crossed the threshold.
Our language just did.
— Edwin O. Paña
We Gather Light to Scatter.
Reflections may be shared beyond this page.
