The Seat at the Table: Artificial Intelligence, Power and the Moral Superiority Trap

Every age believes it is witnessing the greatest technological revolution in history.

But across the long arc of human existence, only a handful of transformations have genuinely altered the trajectory of civilization. The Enlightenment gave humanity a new relationship with knowledge, reason, and science. The Industrial Age transformed production, expanded human capability, and changed the material conditions of life for millions. The computer and the internet connected us in ways once unimaginable, accelerating communication, creativity, and commerce on a global scale. Social media was another such turning point — it brought us closer together, and then it tore us apart.

These were not merely inventions. They were hinges.

And in every case, we built because we could. We never seriously asked whether we should. By the time we got around to the question, the answer no longer mattered.

Artificial Intelligence belongs to that list. But it is also something else.

For the first time, we are not simply creating tools that extend human capability. We are creating systems that operate with increasing autonomy — that make decisions, pursue objectives, and shape the world in ways their creators do not fully understand and cannot fully predict.

This is the first technology capable of participating in the evolution of intelligence itself, ultimately without human involvement.

For most of history, intelligence advanced at nature’s pace. Then we learned to move faster than our understanding could keep pace with. Our tools began to exponentially outpace our moral and ethical evolution, our wisdom. Our capacity to manipulate the world outran our capacity to understand ourselves. We learned to reshape the world faster than the world could reshape us — and we reshaped it to serve exactly one species.

Artificial intelligence will be the hinge we do not see coming.

The question before us is not whether AI will become intelligent. It already is, in ways both narrow and increasingly general. The real question is whether our moral development will begin to keep pace with what we are creating.

Not our intelligence. Not our technological progress. Our consciousness.

Because they are not the same thing.

Intelligence solves equations, optimizes markets, writes code, predicts protein structures, and builds astonishing machines. Consciousness asks a different question.

To what end?

History has demonstrated, again and again, that intelligence without wisdom is capable of extraordinary achievement and extraordinary destruction, usually in very rapid fashion.

Which Brings Us to Silicon Valley

The prevailing philosophy of the technology industry has long been simple: move fast, break things, disrupt, scale.

Those principles produced remarkable innovation. They also produced fractured institutions, surveillance capitalism, engineered addiction, the collapse of local journalism, and an economy concentrated in the hands of a few companies whose influence now rivals that of some modern governments.

Those same incentives are now driving the most consequential technology our species has ever built.

So let us concede the point that AI's builders most want conceded: this was coming.

If not Anthropic, then OpenAI. If not OpenAI, then Google. If not California, then Shenzhen. The transformer architecture was published openly in 2017. The scaling laws were never a secret. Someone was going to press the accelerator.

Grant it. Grant all of it.

And then look closely at what that concession is being used to justify.

An Argument That Cannot Lose.

There is a claim that has become nearly axiomatic inside Anthropic and among many AI safety advocates. State it fairly, because a caricature proves nothing.

If frontier AI is going to exist regardless, then meaningful safety work can only be done by those building frontier systems. You cannot align what you cannot build. Therefore the ethical imperative is not to stand outside the race but to win it — to build the most capable models, raise the necessary capital, earn a seat at the table, and shape the outcome from within.

There is real truth in this. Anthropic has produced serious safety research. Its work on interpretability has advanced the field. I have little doubt that many inside the company are sincere in their concern for humanity.

But sincerity does not exempt an idea from scrutiny. Actually genuine sincerity demands it.

Look at the structure of the argument rather than its conclusion. It cannot be disproven. There is no outcome that would count against it. Every advance in capability becomes an act of responsibility, because had they not built it, someone less careful would have. Every commercial victory becomes a moral victory, because market share is influence and influence is safety.

And in the entire history of the company, this reasoning has never once arrived at a conclusion that conflicted with its commercial interest.

Not once.

That does not prove the argument wrong. It may be substantially correct. But it should stop us cold.

A moral principle that never requires sacrifice — that never leaves money on the table, never closes a door, never asks anyone to surrender power rather than accumulate it — has stopped functioning as a principle. It is a justification.

The claim is familiar: we had to build a multi-billion-dollar frontier AI company, because we needed a seat at the table.

History echoes with versions of it.

Someone was going to build the hydrogen bomb.

Someone was going to extract the resources.

Someone was going to expand the empire.

Better us than them.

It is one of the oldest moral narratives we have, and its distinguishing feature is that it converts the pursuit of power into the acceptance of duty. Power is no longer sought. It is portrayed as an obligation imposed by circumstance.

The people making this argument may believe it completely.

That is precisely why it deserves examination.

Consider what is actually at stake.

I put the odds at even. Half a chance this technology is what saves us from ourselves — the modeling capacity to navigate ecological collapse, the medical acceleration to end diseases we have carried for millennia, the cognitive leverage to finally think at the scale of our own problems. Half a chance it is what finishes us, whether through catastrophe, through the quiet concentration of all meaningful power into very few hands, or through something we cannot yet imagine because we are not yet smart enough to imagine it.

Here is what should unsettle you.

Dario Amodei does not really disagree.

He has publicly estimated something in the neighborhood of a one-in-four chance that this goes very badly. He has also written at length about compressed decades — a century of biological progress delivered in ten years. He holds both at once, the sublime and the annihilating, and he is not being incoherent. He is being accurate. That is genuinely what the distribution of the technology looks like.

But notice what he does with that belief.

A man who believes there is a meaningful chance his work ends the species has concluded that the correct response is to build it. Better funded. Better staffed. Faster.

Not to stop. Not to organize a moratorium. Not to spend his considerable credibility demanding that the thing he says he fears be slowed by force of law.

To build it.

And he is not alone in it. Not because the people building AI are uniquely malicious. Because they are profoundly, recognizably human.

He’s caught in the Moral Superiority Trap.

We are asked to believe that unprecedented concentrations of capital, computation, and influence are justified because the people accumulating them possess superior ethical judgment.

That is not a technological argument. It is a moral one, wearing a technological argument's clothes.

And civilization has heard it before.

Every era produces people convinced they are uniquely qualified to guide humanity. Kings ruled by divine right. Clergy claimed privileged access to truth. Ideologues promised salvation through the state. Each was certain that their superior understanding conferred the authority to decide on everyone else's behalf.

Today the claim arrives in a new dialect — a blend of prosperity gospel and venture capital, in which wealth is not merely the reward for being right but the evidence of it. The vocabulary has changed. The temptation has not.

The greatest moral superiority traps are never built on cynicism. Cynicism can be exposed. They are built on the quiet conviction that because we understand the risks better than others, we have earned the right to decide for all.

I am more ethical than the alternative. Therefore my judgment supersedes yours.

That is the trap. And the deeper danger is not artificial intelligence. It is our enduring habit of confusing intelligence with wisdom, capability with legitimacy, wealth with having intelligence and our own brand of morality with the right to determine humanity's course.

What will it cost us?

I am not going to end by demanding that Anthropic shut down. That demand is built to be refused, and refusing it is free.

Here is what would not be free.

Say the commercial part out loud. We are a company. We are competing. We want to win. We believe our winning is better than the alternatives, and we may be wrong. Every word of that is already true. Saying it costs only the halo — and the halo is the thing actually being defended.

Externalize the referee. Not voluntary commitments and self-administered evaluations. A body with independent authority, adversarial incentives, and the power to say no and be obeyed. Advocate for it in a form that would bind you.

Name a line in advance. Publish, before the fact, a specific circumstance under which you would decline to ship a model a competitor would ship, absorb the loss, and be seen to absorb it. A safety commitment that has never cost anything has never been truly tested.

And retire “we would fall behind” as a core safety argument. It is a business argument. It may be a good one. It is not a safety argument, and the conflation of the two is the beam holding up the entire structure.

Because here is the thing about a form of reasoning that always arrives at keep building. Eventually it will meet a question it should not survive. It will confront evidence it does not want. It will encounter, somewhere in the systems being built, something that ought to stop a serious person cold — and it will produce a research program, a thoughtful publication, a work stream, and the next model.

It will never produce stop.

Artificial intelligence is not only forcing us to confront what machines may become. It is forcing us to confront what we are and what we will become — and whether our evolution in consciousness, humility, and reverence can keep pace with the intelligence we are now bringing into the world.

The odds are not really about the technology.

They never were.