Whoever Wins, Wins: Trump, Obama, and the AI Dividend Nobody's Actually Proposing

On September 14, 2026, President Trump posted on Truth Social that "the only control or 'guardrails' that AI needs is a strong and smart (High IQ!) President," dismissing growing AI safety concerns — including from Anthropic CEO Dario Amodei — as a "sick conspiracy" pushed by China. It capped a fast-moving 48 hours: two days earlier, at a Democratic fundraiser in Manhattan, Barack Obama told House Minority Leader Hakeem Jeffries and assembled donors that AI "is moving very fast in private hands," and urged Democrats to build "a very clear plan" and make it a "central agenda" before the 2028 election cycle. Trump's Truth Social post is widely read by political reporters as a direct rebuttal to Obama's intervention.

Behind the political theater is a genuinely serious infrastructure story: NERC, the organization that oversees the reliability of the North American power grid, issued only its third-ever Level 3 alert after AI data centers caused over 1 gigawatt of simultaneous load loss in Virginia and Texas. PJM Interconnection, the largest grid operator in the country, had a capacity shortfall for 2028 equivalent to roughly seven nuclear reactors, ran an emergency power auction, and is projecting a $6.3 billion increase in consumer electricity costs over the next three years "mostly attributable to data center demand." Capacity prices in PJM's territory have risen roughly tenfold.

Public trust hasn't kept pace with the buildout. A Gallup poll from May 2026 found 39% of Americans now believe AI does more harm than good — up from 31% the year before, and sharpest among 18-to-29-year-olds, where it jumped 11 points in a single year to 47%. A CNBC/Generational Lab poll found at least 65% of young adults don't trust any of nine major tech CEOs on AI, from Elon Musk to Sam Altman to Anthropic's own Amodei.

Every actor in this story is offering a version of the same non-answer. Trump's answer is "trust me, and don't ask questions" — guardrails replaced by a personality, safety concerns rebranded as disloyalty to the country. The tech billionaires' answer is "trust us, we're already thinking about it" — Bezos proposing tax relief that costs him nothing, Musk proposing cash payments funded by a growth model that requires you to also trust his math on inflation, Altman proposing "universal basic compute" instead of money, which is another way of saying: you get access, we keep the equity. And even Obama's answer, as serious as it is compared to the alternative, is still a process — "a framework for a public conversation" — not a mechanism, not a number, not a guarantee.

Meanwhile, the actual infrastructure of control is being built in real time, and it isn't subtle. DHS is pursuing AI-powered smart glasses for ICE agents, built on Meta hardware. Palantir's "ELITE" app for ICE generates deportation-target dossiers with confidence scores attached to human beings. There's a documented case of a woman misidentified twice by facial recognition software and taken into ICE custody because of it. Four Democratic senators had to introduce a bill just to ban ICE and CBP from using facial recognition at all. None of that is a conspiracy theory. It's public record, with the contracts and the bill numbers attached.

The institutional machinery here isn't complicated once you see it: whoever builds the technology gets to write the rules for how its gains — and its risks — get distributed, unless the public forces a different arrangement before that becomes permanent. Right now, on the money side, every proposal on the table is either voluntary (billionaires deciding what to offer), contingent (Alex Bores's actual congressional AI-dividend framework only triggers "if" labor participation declines or wages compress — after the damage, not before), or purely rhetorical (a "framework for a conversation" is not a dividend). On the control side, the infrastructure for surveillance is already being built and deployed faster than the debate about whether it should exist. That asymmetry — fast deployment of control, slow-if-ever deployment of shared benefit — is the pattern. It is not an accident. It is the default outcome whenever the people writing the rules are the same people who profit from them.

AI is not going anywhere, and pretending otherwise wastes the only leverage we actually have — the argument over who owns what it produces, made now, before the terms are locked in. Alaska proved a real dividend is possible almost fifty years ago with oil: 25% of lease revenue, paid to every resident since 1982, cutting poverty an estimated 20-40% without pushing anyone out of the workforce. Nobody currently holding power, in either party, has proposed anything that guaranteed for AI. That's not a technical limitation. That's a choice being made for us, quietly, while we're being told to worry about whether AI has too many guardrails or too few. Notice the pattern. Interrupt it.