When companies replace employees with software, their wage bills shrink. But the taxes that once came from those paychecks cannot simply be transferred to a server.

Imagine an American company removing 100 employees from its payroll after introducing an artificial intelligence system.

Each employee earned $80,000 a year.

The company has eliminated $8 million in annual wages. Management may report the saving as an improvement in efficiency. Investors might welcome the decision.

I would want to examine another part of the accounts.

Those salaries generated tax revenue. Workers paid federal income tax and contributed to Social Security. Their employer also paid payroll taxes.

The employees bought groceries and paid their household bills. Some supported local businesses that depended on their spending.

What happens to those economic contributions when the salaries disappear?

America’s tax system relies heavily on income earned through work. Artificial intelligence may change where that income goes, even if it eventually makes the economy more productive.

The question is whether the tax system can follow the money.

A Lost Salary Means More Than Lower Payroll

An employee earning $80,000 does not take home the entire amount.

The employer withholds federal income tax and deducts Social Security and Medicare contributions.

The company also pays its own share.

According to the Internal Revenue Service, employees and employers each pay 6.2% toward Social Security and 1.45% toward Medicare.

The ordinary combined payroll tax rate is 15.3%.

For 2026, Social Security taxes apply to annual earnings up to $184,500. Medicare taxes have no equivalent wage ceiling.

Return to our imaginary company.

Removing 100 positions paying $80,000 each eliminates $8 million in annual wages.

Those wages previously generated approximately $1.224 million in ordinary employee and employer payroll contributions.

Of that amount, $992,000 went toward Social Security. Another $232,000 went toward Medicare.

These are gross figures based on the hypothetical payroll. They do not establish that the government will permanently lose the full amount.

Some employees may find new jobs. The company may hire workers elsewhere.

But until the lost employment income is replaced, the payroll tax contributions associated with those jobs no longer come from those positions.

A company can save money immediately. The government cannot assume that equivalent contributions will appear somewhere else.

Why America’s Tax System Is Vulnerable

The United States collects federal revenue from several sources.

Individual income taxes provide the largest share. Payroll taxes finance Social Security and part of Medicare.

Corporate income taxes form a much smaller part of the federal tax base.

In October 2026, the Tax Policy Center reported that individual income and payroll taxes together generate roughly 85% of federal tax receipts.

There is an important distinction here.

Not all individual income tax comes from wages. People also pay income taxes on certain investment earnings and business income.

Still, wages and salaries account for a substantial share of the income on which America collects taxes.

Artificial intelligence could gradually change that relationship.

If companies replace some employees with software, less income may flow directly to workers.

A greater share may go to business owners and investors.

That money can still be taxed. But capital income and employment income face different rules.

The timing of tax collections may also change.

An employee pays payroll taxes as wages are earned. A shareholder may not owe capital gains tax until an investment is sold at a profit.

This difference is easy to overlook when executives discuss automation.

Company Profits Cannot Automatically Replace Social Security Contributions

Suppose the company saves $8 million in wages and spends $3 million operating its new AI systems.

It has $5 million left before accounting for other financial effects.

Perhaps that saving increases its taxable profit.

The government may collect additional corporate income tax. Shareholders might also owe taxes when they receive dividends or realize gains.

But corporate income tax does not automatically flow into the Social Security trust funds.

Those funds have their own financing arrangements.

The 2026 Social Security Trustees Report estimates that the Old-Age and Survivors Insurance Trust Fund will exhaust its reserves in the fourth quarter of 2032.

Under current law, continuing income would then cover approximately 78% of scheduled benefits.

AI did not create that financial problem.

The pressure comes from the program’s existing financing structure and long-term demographic changes.

However, a future decline in taxable wages could make the calculations more difficult.

I find this more significant than the familiar argument that robots do not pay taxes.

Machines do not have wages, but the businesses operating them can generate taxable income.

The difficulty lies in collecting enough of that income through the right channels.

Social Security cannot rely on the assumption that higher corporate profits will automatically replace lost payroll contributions.

Are American Workers Already Losing Jobs to AI?

There is a temptation to treat every technology-related layoff as proof that AI is destroying employment.

The research does not support such a sweeping conclusion.

In August 2026, researchers at Stanford’s Digital Economy Lab reported a growing employment gap among younger workers in occupations highly exposed to AI.

Employment among workers aged 22 to 25 in those occupations was approximately 19% below the level it would have reached had it kept pace with employment among similarly aged workers in less-exposed occupations.

That does not mean AI caused 19% of young workers to lose their jobs.

The comparison measures a relative employment shortfall.

The researchers found that reduced hiring appeared more important than dismissals. They also warned that their findings could not establish AI as the sole cause.

An updated Yale Budget Lab analysis, published on October 14, 2026, found no clear evidence of large-scale labor-market disruption associated with AI.

The two studies describe a situation that deserves attention without justifying predictions of mass unemployment.

Some young workers may already be facing fewer opportunities.

At the same time, the American economy has not experienced the nationwide AI displacement that some commentators describe.

AI Could Increase Tax Revenue and Still Create a Problem

Here is the part of the argument that I think deserves more discussion.

Artificial intelligence might increase government revenue.

If businesses become more productive, the economy can grow. Higher output could generate additional taxable income.

That would be welcome.

But the amount collected would depend partly on who receives the gains.

In a study published in July and revised in October 2026, the Yale Budget Lab examined possible AI growth scenarios.

In one rapid-growth scenario, federal revenues in 2030 could be as much as $246 billion higher than the Congressional Budget Office baseline.

That figure is a modelled outcome, not a forecast of revenue that AI will definitely generate.

The researchers found something else.

If income shifted more heavily toward capital rather than workers, the government would collect considerably less revenue than it would under the same economic growth with the existing income shares preserved.

Under one comparison, potential revenue gains were nearly twice as large when the shares of income going to labor and capital remained unchanged.

The lesson is not that AI must reduce total tax revenue.

An expanding economy can generate more tax receipts even while workers receive a smaller share of its income.

The concern is that the government may collect less than the economy’s additional output would otherwise allow.

That is a different problem from the one described in many angry speeches about AI layoffs.

And it may be harder to notice.

Does America’s Tax Code Encourage Companies to Replace Workers?

Economists have also examined whether the tax system influences decisions about automation.

In October 2026, the MIT Stone Center published an analysis based on research by Daron Acemoglu and his colleagues.

The researchers argued that differences between the effective taxation of labor and capital can encourage companies to automate work even when the technology offers limited productivity advantages.

A business pays taxes associated with employment. It can also claim deductions for qualifying investments in equipment and software.

Those deductions may affect the cost of automation.

The tax rules do not mean that machines are untaxed or that replacing a worker always produces a tax advantage.

Business circumstances matter.

Still, the research identifies a policy issue.

A company should have an incentive to use technology when it genuinely improves production.

It is less clear why tax rules should encourage the replacement of employees when the economic advantage is small.

The distinction matters because artificial intelligence can also help people perform their existing jobs.

An accountant who uses AI to examine transactions more efficiently may become more productive without losing employment.

That is different from replacing the accountant altogether.

Would an AI Tax Solve the Problem?

Some researchers have proposed taxing companies when they automate work.

The idea sounds simple.

If technology replaces taxable employment, companies could pay an additional tax.

But how would the government decide when a job had actually been replaced?

Consider an employee who uses AI to finish a report in two hours instead of four.

The company has saved time, but the employee still has a job.

Perhaps the employee now handles more work. Or perhaps the company eventually reduces hiring because existing workers can do more.

An automation tax would need to distinguish between these outcomes.

That would create administrative problems.

The Tax Policy Center’s October 2026 analysis examined possible taxes on AI development and use. It also considered taxes on unusually high profits and automation itself.

Each proposal has disadvantages.

A tax on AI use might discourage applications that help workers become more productive.

A tax on company profits could raise revenue, but defining and measuring excess profits presents its own difficulties.

Broader reform of capital income taxation may offer another approach.

Such reform would not require lawmakers to decide whether every new computer system qualifies as AI.

It would focus on where income accumulates and how that income is taxed.

There are trade-offs here too. Tax rules can influence investment and the location of business activity.

America needs a better answer than simply declaring that every robot should pay the salary tax of the employee it replaces.

The Worker Still Matters After the Payroll Entry Disappears

I have written before about Americans who feel trapped in jobs because they depend on employer-provided health insurance.

In Why 23 Million Americans Feel They Cannot Leave Their Jobs, I examined how health coverage can influence employment decisions.

Losing a job in America can affect much more than salary.

A family’s ability to pay its bills may change quickly. Healthcare costs can add pressure when employment ends.

Even families with substantial incomes struggle with expenses. My article, Why a $100,000 Salary No Longer Guarantees a Comfortable American Life, examined how household costs can leave little room for savings.

AI has the potential to improve that situation.

More productive businesses may offer cheaper services. New industries may create employment.

But the benefits will depend partly on how the economy distributes the gains.

An employee receives wages and contributes payroll taxes.

A company replacing that employee may earn greater profits. The tax consequences depend on the business’s financial arrangements and the laws governing those profits.

Those are different revenue streams.

I am not convinced that AI will destroy America’s tax base. The evidence does not establish such an outcome.

The warning is more specific.

If technology gradually moves income away from workers, governments may have to reconsider how they finance programs built around employment.

Businesses can reduce their payroll expenses when technology changes.

Congress still has to finance Social Security.

The payroll entry may disappear. The public obligation remains.