Universal basic income is a regular cash payment, equal and unconditional, sent to every person as a floor beneath them. Joe Biden’s former commerce secretary Gina Raimondo has decided that UBI in response to AI is the thing to fear most. At the Reagan National Economic Forum she said: "I personally think it's like the end of America."
She is now heading up a newly-launched extremely well-funded organization to make sure the country reaches for anything but a basic income in response to AI. It's called RAISE US, and the money behind it comes largely from the companies building the disruption it exists to manage. Let me walk you through what this new org is, what it quietly concedes, and what it's really for.
What RAISE US Is
RAISE US launched June 25, with Raimondo as CEO and former Indiana governor Eric Holcomb as co-chair. It has already secured more than $500 million toward a $1 billion target. Amazon, Anthropic, Microsoft, and the OpenAI Foundation are anchor partners. Behind them sits a wall of corporate and philanthropic money: Blackstone, General Motors, IBM, Eli Lilly, Mastercard, Deloitte, UPS, Cisco, Workday, Arnold Ventures, and a dozen more.
Their strategy runs through the state level. Arkansas, Connecticut, Maryland, and Utah are first, and on July 31 the National Governors Association announced a partnership with RAISE US to spread the model further. What will the states be testing? Wage insurance. Incentives for companies to retrain rather than lay off. AI career coaching. Redeployment to new jobs.
Every item on that list is a perceived substitute for an unconditional income floor. This is the establishment's carefully assembled answer to a single question: what do we do about AI-driven job displacement that isn't giving people money to directly share the wealth?
The Concession Hiding Inside the Launch
When Amazon, Microsoft, OpenAI, and Anthropic pool half a billion dollars into a state-by-state workforce nonprofit, they are conceding that the labor market will need policy buffers — that the disruption their own technology creates is structural. Nobody raises a billion dollars to smooth over a problem they expect the market to fix on its own. Raimondo herself put it starkly at the launch: if we build the best AI in the world and leave millions of Americans behind, we won't have won anything, we'll have automated our own decline.
Great. We agree. The disruption is real, it's coming, and the existing safety net won't save us.
So what's the plan? Retraining?? The one response we have already studied for sixty years.
If that’s not funny to you, this part’s even funnier. Raimondo has herself called past federal retraining efforts "ineffective." In a New York Times op-ed ahead of the launch, she wrote that an AI economy crisis is coming and there's "no obvious solution." The woman raising a billion dollars for retraining has told us retraining doesn't work and that she sees no obvious solution — while the solution with the strongest contemporary evidence base sits right in front of her, and she calls it the end of America.
Retraining Doesn’t Work the Way They Need It To Work
The most comprehensive synthesis we have is a meta-analysis of more than 200 retraining program evaluations worldwide. Its finding? The average training program does close to nothing in the first year, and even two to three years out the payoff is modest. The authors size it at roughly the bump from finishing one community college course.
Now consider the price tag. Trade Adjustment Assistance, the federal government's flagship program for displaced workers, got the gold-standard treatment: a nine-year Mathematica evaluation matching participants against similar laid-off workers. Employment and earnings effects came out essentially neutral, with TAA participants actually earning less than the comparison group. The cost-benefit accounting found a net loss to society of roughly $54,000 per participant. That is what it costs to spend a fortune making people slightly worse off.
There is one real exception: sectoral training, if built backward from concrete, identified demand, with employers actively waiting to hire. Randomized trials of Year Up, Per Scholas, and Project QUEST show large earnings gains that persist more than a decade out. Project QUEST trains people for health care, exactly the work an aging country needs more of. We should run a federal sectoral program for care work tomorrow. It's a good thing, and we already know it helps.
But look at the condition that makes it work: the jobs must be real, needed, and wanted. That's a program to run alongside a universal basic income. It is no answer to AI on its own.
The Newest Study Is the Most Damning of All
If you don’t already think it’s ludicrous to lean into job retraining after a decade of telling people to learn to code, get a load of this next bit.
In May, four economists published an NBER working paper asking the correct question: how retrainable actually are AI-exposed workers? They assembled 1.9 million federal training records from 2012 to 2024, linked them to AI exposure measures, and compared trainees against matched workers who got only job search help. Workers from highly AI-exposed jobs earned about $900 more per quarter in the trainee group than the job search group before 2020, rising to roughly $2,900 by 2022–2024. They estimate that only 40% to 45% of occupations are "AI retrainable."
That's the best news retraining has gotten in years, right? Now read how those workers won.
The gains for AI-exposed workers came primarily from moving into less AI-exposed work. Not from learning to work alongside the machines. From running away from them. The paper's own table of the most common transitions during the AI boom tells a story of how customer service representatives became heavy truck drivers and nursing assistants. Shipping clerks became truck drivers too. The authors call it, in their words, an AI-avoidance narrative.
So the good news is that retraining works, provided we define "works" as climbing down the ladder into physical labor the machines haven't reached — yet.
We spent fifty years building a system to move people up the ladder into knowledge work. The best available evidence now says its main function in the AI era should be to move them back down, and we're calling that a solution.
And how long does that escape route stay open? The destination occupation in study after study is truck driver. Ask yourself what autonomous vehicles do to that plan. Driverless trucks are already arriving. We are training displaced office workers into the exact jobs next in line. This is an exit route to an island that's sinking.
Three more things about that paper: The workers in it averaged about $40,000 a year before training. That’s a program for the working poor, not for the forty-five-year-old attorney or the laid-off software engineer, and no part of our system was built for them. The authors note returns are higher when labor markets are tight. But if AI produces a slack market due to mass unemployment, that advantage disappears.
That's the strongest case for retraining that exists. Yikes.
Now Look at the Evidence They’re Ignoring
RAISE US says its purpose is to find what works state by state and scale it up. Okay. Cool. Then look at what's already been tested, over and over, in state after state.
A guaranteed basic income floor has been studied in over two hundred American pilots. Not one has shown a worrying collapse in work. Many show employment increases. People use the money to leave a survival job for a better-matched one. Nineteen-year-olds use it to choose school over the first paycheck they can find, which is to say, they invest in exactly the skills RAISE US claims to want them to have.
The closest thing we have to a meta-analysis of UBI experiments is a 2020 peer-reviewed systematic review that screened more than 1,200 cash studies using PRISMA — the same method medical researchers use to weigh drug trials — hunting specifically for the thing those like Raimondo fear. Their fears are not supported. Labor supply actually rises: among men and women, young and old. The only places work falls are the places we should want it to. Children. The sick and disabled. Parents of young kids. Young people who stay in school longer. All of it more than offset by a larger number of workers who find new jobs due to greater consumer spending.
So here we have one approach with sixty years of mediocre results that Raimondo herself calls ineffective, and one with decades of hundreds of pilot experiments showing it does considerable good across a wide range of measures without any worrisome impact on work. RAISE US is looking to pour a billion dollars into the first while treating the second as the end of America. They say they want what works, but they're ignoring what works.
We Ban What Threatens Us, Not What Fails
Basic income pilots have been so successful that people who agree with Raimondo have moved to ban them. Iowa outlawed local guaranteed income programs in 2024, killing the UpLift pilot — a program whose results were positive, where the money went to food and housing for families with kids. South Dakota, Idaho, Arkansas, and Texas have done the same or tried. Arkansas, I'd note as a state that banned UBI programs, is RAISE US's first partner state.
Think about that for a minute. Nobody bans a program because it fails. The evidence of failure speaks for itself. You ban it because it works, and the political implications terrify you. No one has ever introduced a bill to ban job retraining, and no one will, even though most all retraining programs fail and plenty of the ones RAISE US funds will fail too. Failure takes care of itself. We only outlaw what threatens our evidence-averse ideologies.
What This Is Actually About
So why retraining? Why pour a billion dollars into the thing we know underperforms while treating the thing that works well as a national emergency to prevent?
Because retraining keeps the power to say No out of workers' hands.
That's the real issue. This was never about UBI making people "give up on work." A basic income lets a person say No — no to the first lousy offer, no to the bad hours, the bad pay, the bad boss, the office that treats people like inventory — and hold out for the job that fits them and their skills, pay, time, and location preferences. Retraining keeps you dependent on the next employer. A floor lets you wait for the right one offering the right job. One of those things is freedom. The other is a leash with a training certificate attached.
Raimondo's own words give it away. In April she called UBI a "cop out" and "bad policy," and said: "You want social upheaval? Try that. Everybody stays home doing nothing all day and gets paid by the government." That’s not following the evidence. It's pure ideology wearing a policy costume. It’s the belief that people must be kept needing to work for the owners of survival resources in order to survive, regardless of the amount of sheer abundance. It’s the belief that true freedom is dangerous, because it includes the freedom to do what we want instead of what those like Raimondo want for us.
UBI Is Bigger Than Job Loss Anyway
Raimondo's whole frame misses the point of UBI in response to AI regardless.
A basic income is not merely a parachute for people AI lays off. The reason the floor is universal — beyond making sure nobody falls through it — is to spread the gains of AI to everyone, like the productivity dividend it should be seen as. We all trained these large models, with our work and our words. If you're a worker whose job AI never touches, retraining does not compensate you for your training data. A dividend pays you anyway, as a shareholder in the technology you helped build along with the rest of us.
It also keeps the economy on solid ground. For fifty years productivity climbed and wages didn't, and the gains flowed to the top. AI widens that gap. Not one dollar of RAISE US retraining touches it. A floor does. When someone loses a $100,000 job, a $1,500 monthly basic income is an amount of money they keep spending — and their spending is somebody else's paycheck. That's how you break the deflationary spiral where one lost job kills the next. In an economy that runs on consumer spending, the real job creators are consumers. Fewer than one in three unemployed workers even qualifies for unemployment insurance. A basic income reaches three out of three. It's the only floor that never lets anyone hit zero, and the ground under every entrepreneur, every caregiver, every person who'd absolutely build something if they weren't one paycheck from ruin.
Retraining doesn’t do squat to create jobs. People spending money on stuff creates jobs.
I Lived Through “Learn To Code”
Now I want to mention how Paul Ryan sits on the RAISE US advisory board.
Ryan represented Janesville, Wisconsin, where General Motors padlocked their plant two days before Christmas in 2008. Janesville did everything that was supposed to work. Laid-off workers went back to school. The technical college steered them toward the fields the labor data called hot. And the people who retrained ended up no better off than the neighbors who didn't, and by most measures worse. Ryan spent 2012 empty promising from behind podiums to retrain the displaced for the jobs of the future. Fourteen years later he's on the advisory board of a billion-dollar organization making the same promise about AI.
David Autor is on that board too — the economist whose own work documented that China Shock towns never recovered from the offshoring of jobs. Those jobs simply never came back. But sure, this time, job retraining will work.
I lived through the "learn to code" years. I said over and over that retraining wouldn't save the coal miners we were turning into coders. Now the coders are being replaced by code, and the answer on offer is to retrain them into plumbers and electricians. It's the Arrested Development bit made policy: it never works, but maybe it'll work this time, for us.
It won't. RAISE US is an anti-UBI containment strategy dressed as a workforce initiative; a mountain of nonprofit money spent to delay the one thing that would actually give people real freedom of choice. The people funding it see those choices as dangerous. They are not offering freedom. They've decided freedom is the threat, and they've built an organization to hold it off.
We should have started a universal basic income fifty years ago, when wages first decoupled from productivity. We didn't. We cannot afford to lose another generation while billions of dollars are set on fire proving, one more time, what we already know.
