Jesse Marks · Substack · 25 Jun 2026 essay
AI Overconcentration and the End of the Balance of Power?Applies balance-of-power theory from international relations to frontier AI control.
Is it acceptable for a small number of private companies to steward systems with potentially society-wide influence?
The concern now runs beyond market power: analyses of AI-enabled coups argue that whoever controls advanced AI could convert it into political power directly — making internal lab governance a constitutional question.
View on the map → · Open in Browse →
A new axis opened: after Washington forced frontier models off the market, the contest is labs versus the executive branch, not just labs versus everyone else. Cotra argues a company could rival states outright; the OpenAI-restructuring advocates moved to SEC disclosure; a funder now treats extreme concentration as a research field. The discomfort went mainstream: Krugman argues AI's shock is landing on an already-oligarchic wealth distribution, and Amodei conceded on the record being 'somewhat uncomfortable' with a concentration of power arriving 'almost overnight, almost by accident.'
Unchecked control of frontier AI is a systemic, possibly irreversible risk.
Applies balance-of-power theory from international relations to frontier AI control.
Private companies should not have the power of nation states.
Argues a leading lab could accumulate state-rivaling power during rapid takeoff; proposes external model access and law-following mandates.
If transaction cost is 'train your own frontier language model'...firm size = everybody
A Coasean mechanism for concentration: frontier-model economics push toward firms of unprecedented scale.
The current concentration of wealth at the top is not only immensely greater than in the 50s or 60s, but is in fact far greater than during the Gilded Age.
Krugman argues AI's shock is landing on a pre-existing oligarchic wealth distribution, so its harms will be amplified relative to a counterfactual egalitarian economy — a prominent economist joining the concentration debate.
I'm at least somewhat uncomfortable with the amount of concentration of power that's happening here. I would say almost overnight, almost by accident.
A frontier-lab CEO conceding on the record that the current concentration of power in a few AI companies is accidental and uncomfortable — a notable admission from inside the concern.
What matters is power answering to competing institutions, not power being small.
Freedom has never depended on power being small. It has depended on power being answerable to more than one authority at a time
Concrete mechanisms proposed or funded.
From the advocates who secured California oversight of OpenAI's restructuring — securities disclosure as the next accountability lever.
Treats extreme power concentration as a fundable research field and directs money at mitigations.
It's a one-way door and we weren't ready for it.
A new axis the debate lacked: control contested between labs and an executive branch, after the US government forced frontier models off the market.