Cyber Nexus Logo

Cyber Nexus

Back to home

Web8 min read

How AI Could Change Jobs—and What Sam Altman Is Proposing

If AI helps you finish a day's work sooner, who gets the saved time? A look at changing jobs, the risk to entry-level work, and Altman's ideas for sharing the wealth—from an equity fund to universal basic compute.

How AI Could Change Jobs—and What Sam Altman Is Proposing

Imagine a support team that starts using AI to draft replies. The queue gets shorter, and each person can handle more requests. Does the company improve its service, stop hiring, cut the team, or let people finish earlier? The tool doesn't make that decision. Management does. That's why the future-of-jobs argument can't be settled by showing that a model can complete a task.

Sam Altman has spent years arguing that AI could create enormous wealth while shifting power away from people who earn a living through work. His proposed responses are worth examining, but there isn't one finished 'Altman plan'. There are personal essays, an idea about sharing computing resources, and a separate set of policy proposals published by OpenAI. They don't all promise the same thing.

Start with the tasks, not the job title

A developer's job includes writing code, understanding an unclear request, debugging a production failure, reviewing a change, and deciding what shouldn't be built. Automating a first draft changes that mix. It doesn't establish that the whole job can disappear. The same distinction applies to an accountant checking a suspicious invoice or a support agent deciding when a standard refund policy doesn't fit.

The ILO and Poland's NASK made this distinction in their May 2025 occupational-exposure study. They estimated that one in four workers globally was in an occupation with some exposure to generative AI, and identified job transformation as the more likely overall impact. Exposure describes the potential for tasks to change. It is not a count of jobs already lost, or a prediction that a quarter of workers will be laid off. The report is linked in the sources below.

That doesn't make displacement a minor concern. If a business can produce the same amount with fewer people, some roles may go. Lower costs can also create demand for work that previously wasn't affordable. Both are possible, and they needn't happen in the same place, at the same time, or to the same workers.

A harder first step into a career?

One risk I'd watch is what happens to beginner work. Small fixes, routine research, and first drafts aren't just outputs; they're how people learn to handle harder assignments. If a team automates those tasks and recruits only experienced reviewers, where does the next group of experienced people come from? That's a risk to plan for, not proof that every junior role is about to vanish.

Teams could keep supervised projects, pair junior staff with experienced reviewers, and teach people to test an AI-generated answer rather than simply accept it. New work may also grow around integration, evaluation, security, and customer needs. But listing possible new roles doesn't explain how a displaced worker pays the bills while qualifying for one.

Altman's 2021 proposal: give people a stake in the assets

In 'Moore's Law for Everything', published in March 2021, Altman proposed an American Equity Fund. His reasoning was that if AI moves more economic value towards companies and asset owners, people should share in that ownership rather than depend entirely on wages. He also expected technology to make many goods and services cheaper. That's his economic thesis, not a guarantee that rent or healthcare will fall on the same schedule as software costs.

  • The fund would receive an annual tax of 2.5% of the market value of companies above a valuation threshold, paid in shares.
  • It would also receive a 2.5% annual tax on privately held land value, paid in dollars. The proposal distinguishes land from the buildings and improvements on it, while acknowledging assessment difficulties.
  • Eligible adult US citizens would receive annual distributions in cash and company shares. Altman proposed phasing in the tax rates rather than starting at the full rate immediately.

This was a tax-and-ownership proposal, not a promise that OpenAI would pay everyone a salary. It also isn't quite the same as a fixed universal basic income: distributions tied to assets and their value can vary. Funding, valuation, eligibility, and democratic oversight would need to be worked out through public policy.

Basic income and 'basic compute' solve different problems

In a May 2024 All-In podcast discussion, reported by POLITICO, Altman floated universal basic compute: give people a share of AI computing capacity that they could use, resell, or donate. Instead of receiving only money, a person would have access to part of a productive resource. It was an exploratory idea, not a detailed benefit programme with settled rules.

There's an appealing part to that. A student, a small business, or an independent researcher could use computing capacity without already having a large budget. But compute isn't cash. You can't pay a grocery bill with it unless there's a reliable way to turn the allocation into money. Its usefulness would depend on access, transfer rights, pricing, and who controls the service. Cheap access to a model is not the same thing as owning the company behind it.

The optimistic picture—and OpenAI's more concrete 2026 ideas

Altman's June 2025 essay 'The Gentle Singularity' describes a future of much cheaper intelligence, faster discovery, and new things for people to do. It also acknowledges that whole classes of jobs could go away. His proposed direction is to make advanced systems safe and distribute access widely. The optimistic outcome depends on difficult technical and political choices; calling a transition gentle doesn't make it gentle for someone who loses their income.

OpenAI's April 2026 paper, 'Industrial Policy for the Intelligence Age', goes further into possible institutions. This is a company policy document, not another personal Altman essay. Its economic proposals include:

  • A Public Wealth Fund that could invest in diversified long-term assets and distribute returns to citizens. How to seed it is left for policymakers and AI companies to determine.
  • Time-limited pilots of a 32-hour, four-day week with no loss of pay, while maintaining output and service levels. The proposal is to test sharing efficiency gains, not declare every job suitable for a shorter week.
  • Safety nets that respond to measured disruption, including options such as wage insurance, temporary cash support, and training assistance.
  • Benefits that follow people between jobs, plus worker participation in decisions about how AI is deployed.
  • Broader access to useful AI and changes to the tax base that could help fund public programmes as the balance between labour income and capital returns changes.

OpenAI explicitly presents these as early, exploratory proposals, focused on the United States as a starting point. The paper is evidence of what the company advocates, not evidence that those benefits have been delivered. Its Public Wealth Fund also shouldn't be treated as an exact copy of Altman's earlier fund with all the same tax rates attached.

The question I'd ask of any of these plans

Who carries the risk while the promised wealth is arriving? A fund may build assets over time, but someone whose hours are cut needs help now. Investments can fall in value. A US citizen dividend wouldn't automatically support workers elsewhere. And a policy that shares investment returns doesn't, by itself, give employees a say over workloads or redundancies.

Altman leads a company that sells the technology at the centre of this debate. That doesn't invalidate his proposals, but it makes independent evidence, worker representation, competition, and public oversight important. We'd be taking a large risk if the entire social safety net depended on an AI company's forecast turning out to be right.

What to do without betting your career on a prediction

  • Pick one low-risk task in your actual work. Compare doing it with and without AI, including the time spent checking and fixing the output.
  • Keep building subject knowledge. Knowing why an answer is wrong matters more than being able to generate ten plausible versions of it.
  • Use approved tools and respect confidentiality. A faster workflow isn't worth leaking customer records or private company material.
  • If you manage people, preserve learning opportunities and discuss how saved time will be used. Measure quality and workload, not just how many outputs the team produces.

Learning the tools can help, but it isn't insurance against a layoff. People also need choices about training, bargaining, income support, and how the gains are shared. The useful test for an AI jobs plan is whether it improves those choices—not how confidently it predicts a world where nobody needs to worry about work.

Sources and further reading

Sources reviewed on 7 November 2026. The examples of workplace decisions above are illustrative; the dated proposals and research findings come from these publications.