A Verasight survey of 1,690 U.S. adults found that 69% supported requiring AI companies to transfer 50% of their stock to a public fund. That is a striking result for a policy that would give the public not only a share of AI wealth, but potentially enormous voting power over some of America’s most valuable companies.
It is also a result that becomes more interesting—not less—when stated precisely.
The poll sampled adults, not only workers. Respondents saw one of two randomized descriptions. The version that did not name Senator Bernie Sanders received 69% support. When Sanders was identified as the proposal’s author, support fell to 64%. And the one-sentence survey question omitted most of the mechanics that make the actual bill consequential: thresholds, voting shares, corporate separation, governance, and a one-time equity tax.
The correct conclusion is not that Americans read a complex bill and endorsed every clause. It is that a broad public instinct has formed: if AI creates extraordinary private wealth while disrupting work, the public should own part of the upside.
TL;DR: what does the 69% poll actually show?
Question
Direct answer
Who was surveyed?
1,690 U.S. adults age 18+, designed to be nationally representative
What was asked?
Whether respondents support requiring AI companies to transfer 50% of stock to a public fund
Where does 69% come from?
The randomized prompt that did not name Bernie Sanders
Support when Sanders was named?
64%
Was this a worker-only poll?
No; coverage calling it a survey of workers overstates the sample
What bill inspired it?
S. 4825, the American A.I. Sovereign Wealth Fund Act
Does the bill cover every AI user?
No; it targets qualifying large and systemically important AI businesses
“Would you support requiring AI companies to transfer 50% of their stock to a public fund?”
Respondents were randomly assigned to a version that named Sanders or one that did not. The difference was five points:
Prompt
Support
Policy described without sponsor
69%
Same idea identified with Bernie Sanders
64%
In the unnamed condition, support crossed party lines: 82% of Democrats, 77% of independents, and 59% of Republicans backed the idea.
That is real evidence of a broad coalition. But a single high-level question cannot reveal whether respondents support voting control, a 50% one-time tax, direct cash dividends, specific eligibility thresholds, or the fund’s proposed commission. Polling a value—“the public should share AI wealth”—is easier than polling an institution.
There is also a reporting correction worth making. CNBC’s July 12 headline described a majority of U.S. workers supporting the fund. Verasight’s methodology describes a national sample of adults. Some respondents were workers; the reported 69% is not a worker-only estimate.
How strong is the survey methodology?
Verasight says the survey used its verified, multi-mode panel and was weighted to represent the U.S. adult population. The reported sample size was 1,690, with a ±2.8 percentage-point margin of error that accounts for design effect.
Two limitations should travel with the headline:
Randomized-arm sample sizes are not shown. The public report gives the full survey N, but not the base size or response distribution for each version of the public-fund question.
The field dates conflict. The primary report shows June 18–29, 2026, while a syndicated July 2 Verasight press release says June 18–19. We use the dates displayed in the primary report, but the discrepancy should be corrected by the pollster.
None of this makes 69% meaningless. It means the responsible description is: a large national online panel found broad top-line support for an unnamed 50% AI-stock transfer, within a wider survey showing strong demand for AI oversight.
Other questions in the survey support that broader interpretation. Verasight reported 89% support for public disclosure of internal AI-safety tests and 81% support for giving the federal government authority to block systems deemed too risky. The wealth-fund result sits inside a mood of low trust in self-regulation, not as an isolated enthusiasm for public investing.
What Sanders’ bill would actually do
The American A.I. Sovereign Wealth Fund Act, introduced as S. 4825 on June 18, 2026, is much more specific than the poll prompt.
Apply to qualifying or systemically important AI businesses with at least $200 million in relevant annual gross receipts
Cover major AI services, data centers, compute infrastructure, and advanced robotics—not every company using a chatbot
Levy a one-time excise tax paid in stock, equal to 50% of outstanding equity interests
Transfer those shares to a U.S. Treasury sovereign wealth fund
Create a seven-member Independent Commission for Democratic AI, nominated by the president and confirmed by the Senate
Give the fund voting rights and a mandate to influence company decisions in the public interest
Direct an annual distribution equal to 5% of the fund’s market value toward direct payments and public purposes such as housing, education, and health care
Require structural separation when a large mixed business combines covered AI activity with unrelated lines of business
Sanders estimates the initial fund could be worth nearly $7 trillion and pay more than $1,000 annually to every American. Those are sponsor projections, not an independent Congressional Budget Office score.
The 50% figure therefore does not mean a cash tax equal to half of profits or half of market value. It means equity interests—a transfer of economic ownership and corporate power.
Why this idea has escaped the political fringe
Public wealth funds are not new. What is new is the argument that AI itself is a natural resource-like source of economic rent.
AI companies depend on public research, national energy grids, education systems, data created by society, government contracts, and capital markets. If a small number of firms turn those inputs into highly concentrated ownership while reducing demand for labor, advocates argue that the public deserves an automatic claim on the gains.
That argument now appears across ideological and industry lines:
Sanders proposes compulsory 50% public ownership for the largest AI businesses.
President Trump and administration officials have discussed government equity participation, though not Sanders’ structure.
AI executives have increasingly discussed taxes, dividends, or safety nets for technology-driven disruption.
The politics are being pulled by the labor market. May 2026 brought 97,000 announced job cuts with AI frequently cited, while public confidence that productivity gains will reach ordinary workers remains weak. Whether every “AI layoff” attribution is credible is a separate question; the distributional anxiety is already real.
A public wealth fund is ownership, not welfare
The best argument for a fund is that it changes who owns productive capital.
Training helps people compete for jobs. Wage insurance helps after displacement. Universal basic income redistributes cash. A public wealth fund gives citizens a claim on the asset producing the disruption.
That distinction matters in an economy where labor income could grow more slowly than capital income. Our guide to surviving AI job automation focuses on what an individual can control—skills, savings, role design, and AI fluency. A wealth fund addresses what individual preparation cannot solve: a society where gains accrue mainly to model owners, cloud providers, chip companies, and early equity holders.
The idea resembles “universal basic capital”: build a diversified pool of assets first, then distribute a portion of long-term returns. If managed well, the fund can compound instead of relying on a new annual appropriation.
The 50% design creates enormous trade-offs
Supporting shared AI wealth does not automatically make a one-time 50% equity tax the best mechanism.
1. It dilutes existing owners
The shares do not appear from nowhere. Issuing or transferring half of outstanding equity reduces the economic and voting position of founders, employees, venture funds, public pensions, index funds, and future shareholders.
Some affected investors are extremely wealthy. Others are retirement savers. A serious policy analysis needs to trace the incidence rather than treating “AI companies” as a single billionaire.
2. It changes corporate control
A 50% stake with voting rights is not a passive dividend instrument. The public commission could block or shape product, investment, safety, labor, and acquisition decisions. Advocates see democratic accountability; critics see political management of frontier technology.
The governance quality of the commission would become as important as the governance quality of the companies.
3. The threshold creates hard boundary problems
The bill uses revenue and systemic-importance tests, but “AI business” can be difficult to separate from cloud computing, advertising, enterprise software, logistics, and robotics. Structural separation tries to solve that problem by breaking covered AI activity away from other lines.
That can make the tax base cleaner. It can also reorganize some of the largest companies in the economy.
4. It may change investment incentives
If crossing $200 million in AI revenue triggers a 50% equity transfer, firms may delay recognition, split operations, move activity, license rather than sell, or stop just below the threshold. New entrants could face a discontinuity precisely when they begin to scale.
Policy design should tax durable economic rents without making growth itself the taxable event.
5. A concentrated AI fund can fail with the sector
If AI valuations fall, job disruption and fund losses could arrive together. A durable public fund should diversify over time instead of remaining a leveraged bet on the companies it is meant to govern.
6. Sponsor forecasts are not guaranteed returns
The proposed $7 trillion starting value and $1,000+ annual payment depend on valuations, coverage, legal survival, transfer mechanics, and a 5% distribution rate. A bill can specify the payout formula; it cannot specify market returns.
Our take: 69% is a mandate for the principle, not this exact instrument
The survey result should make policymakers take shared AI upside seriously. It should not end the design debate.
The public appears to be saying three things:
AI companies should not write their own rules.
Workers expect genuine economic displacement, even if the timing is uncertain.
Access to AI products is not the same as ownership of AI wealth.
That is a stronger signal than “voters have become socialists” or “nationalization is inevitable.” The unnamed question compresses a complicated institution into a fairness intuition. The Sanders cue reduces support slightly, but a majority remains.
The most robust policy might combine several tools rather than depend on one 50% transfer:
Tool
What it captures
Main weakness
Public equity contribution
Long-term upside and governance
Valuation, dilution, political control
Excess-profit or windfall tax
Realized rents
Volatile revenue; avoidance
Compute or energy levy
Measurable infrastructure scale
Can penalize useful investment
Automation / robot tax
Labor substitution
Hard to define the displaced unit
Data or royalty dividend
Value derived from public inputs
Attribution is technically difficult
Public co-investment
Upside from grants, loans, and infrastructure
Requires capital and investment discipline
Worker ownership requirements
Shares gains inside firms
Does not reach displaced outsiders
A better fund would have five design commitments:
Independent governance with transparent appointments, conflicts, voting records, and audits
Broad diversification after initial AI contributions
Automatic rules for deposits and distributions, limiting election-cycle raids
Portable citizen benefit, not a subsidy tied to one employer or platform
Worker-transition funding for training, wage insurance, mobility, and local communities before mass unemployment is proven
The goal should be to socialize part of extraordinary upside without turning every model release into a political allocation decision.
The deeper shift: AI policy is becoming economic policy
For most of the last decade, AI governance focused on bias, privacy, safety testing, and misuse. Those remain essential. The wealth-fund debate marks a second phase: Who owns the productivity gains?
Anthropic’s Economic Index on automation and worker sentiment shows why that question is difficult. Heavy AI users often feel more optimistic about their own prospects even while worrying about juniors and the broader labor market. Productivity can improve individual bargaining power and increase systemic inequality at the same time.
Likewise, our job transformation map for 2027 identifies new roles without pretending every displaced worker will move into them smoothly. A social wealth fund is attractive because it does not require every citizen to win the reskilling race before sharing in growth.
What to watch next
The poll will matter only if it survives contact with institutional detail. Watch for:
Committee action on S. 4825 and any Congressional Budget Office or Joint Committee on Taxation analysis
Constitutional and tax-law challenges to a compulsory equity transfer
Alternative bills with smaller equity stakes, profit taxes, or public co-investment
OpenAI’s own fund proposal, including whether company contributions are voluntary and how governance would work
Public support after trade-offs are named, such as dilution, government voting power, and possible effects on investment
If support remains high after those details are included, the 69% number will look less like an expressive poll response and more like a durable change in the social contract.
Poll figures and bill status are accurate as of July 13, 2026. Survey support measures responses to short policy descriptions, not approval of every provision in S. 4825. This article is policy analysis, not financial advice.
