Rethinking the Tax Code for an AI Economy

Our federal tax base will be taking a major hit - we need to get ahead of it to prevent the erosion of key social programs.

She seemed unsettled. Jen and I had been friends since ninth grade. I could always sense when something was off, even if she pretended she was fine. We met the first morning of the first day of school. Our lockers were right next to one another's. She shared her removable floral locker wallpaper with me and we became fast friends.

Now, sitting in a downtown cafe years later, sipping lattes as we do our monthly catch-up, I asked her what's wrong. She finally came out with it: she was becoming obsolete. It had happened in a matter of months. She didn't know what it meant for the future of her career and her family.

Jen had risen quickly at the boutique Wall Street investment banking firm and was their star trader by her early thirties. It didn't surprise me–she had always been smart and driven. She had made it to an Ivy League school for college, then an MBA.

The offices surrounding hers, once bustling with phone calls and rapid-fire trading decisions, were quietly replaced by a cluster of servers and lines of code. One by one, traders had been let go as the firm increased its reliance on automation for its trading analysis and recommendations. She was likely next on the chopping block. Those servers didn't have to pay mortgages or send kids to college–they could execute trades with ruthless efficiency and nominal cost.

It is becoming evident that the next great financial innovation will eliminate jobs. In the gleaming towers of Wall Street, lines of code are quietly becoming more valuable than the professionals who once ruled the halls. It is possible that long-term, AI will generate more jobs than it cuts. However, this remains to be seen.

What is certain is that the finance industry's progression towards increased reliance on automation will have implications for tax revenues. Across Wall Street, algorithmic trading and AI-driven portfolio management are increasing efficiency but reducing headcount. Yet highly-paid roles like traders, portfolio managers, and financial analysts contribute substantial income tax and payroll tax.

As AI trading systems and robo-advisors replace some or many these jobs, our government will see lower personal income tax receipts. The government loses that worker's income and payroll taxes. This can potentially amount to hundreds of billions of dollars annually. Our tax code was designed to tax labor-based income, not capital-based income. Once the labor is the capital, the system falls apart. Moreover, financial firms currently enjoy tax advantages for technology investments (e.g. R&D credits or expensing of software), effectively incentivizing automation. This creates a structural revenue drain if the finance workforce shrinks due to AI.

And let's be clear–it's not just the finance industry that will adopt AI and drive erosion of our collective tax base. Finance is one of many industries, including the transportation, medical, and legal industries, that will drive the nosedive in federal tax revenues, our primary source of funding for social safety nets like Medicare and social security. We must modernize the tax code to meet the demands of an AI-driven economy.

U.S. lawmakers are grappling with how to adjust the tax code in response to AI-driven automation. One idea gaining traction is a so-called "robot tax." Bill Gates has proposed that companies pay taxes on robots or AI systems that replace human workers. The logic is that as companies automate jobs, they should contribute more to the social costs of displaced workers and help fund new opportunities. Tech-driven productivity gains should benefit society at large. A robot tax would also enable companies to recoup lost payroll tax revenue.

Bill Gates has argued forcefully for this approach, suggesting that if a robot replaces a human's work, it should be taxed at a similar level to the income taxes that worker was paying. Gates believes this could both slow the pace of automation and inevitable displacement of jobs, giving society time to fund training and service roles for displaced workers, especially in critical areas like eldercare and education. Other prominent voices echo this sentiment. For example, Mark Cuban had to-the-point social media post: "Tax the robots."

Supporters, such as Senator Bernie Sanders, who endorsed Gates' idea in early 2023, also point to inequality concerns, arguing that a robot tax is a tool to redirect some of AI's wealth to workers and remedy inequality if properly used to fund a stronger safety net. New York City Mayor Bill de Blasio proposed automation taxes to require companies to "give back" when they replace employees, using the revenue to train and hire displaced workers. This underscores a broader pro-tax viewpoint: as AI advances, companies should bear greater responsibility for the social impact, either through direct robot taxes or similar levies, to ensure we don't hollow out the tax base that funds public services.

However, the International Federation of Robotics (IFR), which represents robotics manufacturers, strongly opposes such measures, calling Gates' robot tax proposal a solution to "a problem that does not exist." They point to historical data showing that automation often creates as many jobs as it displaces by boosting productivity. For example, as robots were adopted in the U.S. auto industry, human employment in that sector increased, as did productivity gains. The IFR's president, Joe Gemma, argues that "Profits, not the means of making them, should be taxed."

Similarly, many tech executives worry that a robot tax would discourage investment in AI and robotics, potentially pushing companies to offshore automation or lose incentive to innovate, thereby doing more harm than good. They argue that taxing automation is misguided because it risks undermining the very productivity and growth that new technology delivers. The OECD found that companies that employ technological innovation effectively generate up to 10 times more productivity than those that do not. Therefore, slowing AI adoption via taxation could ultimately harm workers in the long term.

A Brookings Institution article by Robert Seamans entitled "Tax not the robots" suggests an alternative fix: if the concern is that automation lets companies earn more while payroll tax revenues fall, the solution could be to raise taxes on capital income or corporate profits rather than taxing the robots themselves. This avoids creating a perverse incentive against innovation. Robert Reich, a prominent public policy professor at UC Berkeley, has also advocated for higher capital gains and corporate taxes as an alternative to robot taxes.

Another approach floated in Congress is a financial transactions tax (FTT) – a small levy on trades of stocks, bonds, or other assets that could both temper ultra-high-frequency AI trading and generate revenue. According to a Brookings Institution study from 2020, the Congressional Budget Office estimated a 0.1% FTT could raise hundreds of billions over a decade. Former tech entrepreneur Andrew Yang has championed this idea, proposing that a 0.1% financial transactions tax could raise as much as $50 billion per year to help fund programs supporting workers displaced by automation, while not materially harming ordinary investors.

The logic is that because trading volumes are so huge, even a 0.1% fee (just $1 per $1,000 traded) on Wall Street's AI trades could generate hundreds of billions over a decade. Proponents note that since 70% of stock market volume is now algorithmic trading, a transactions tax would capture revenue from AI-driven market activity. In essence, FTT is a way to tap into a largely automated profit stream for the public good, while avoiding direct taxation on productive investments like R&D or robots in factories.

There are those in finance and tech that worry that even a small transactions tax could discourage beneficial trading and make it harder for companies to raise capital. High-frequency trading firms might relocate to overseas exchanges to avoid the tax, undermining the policy if not applied globally. Some in the startup and VC industry fear that an FTT could indirectly hurt startup investors by reducing the fluidity of stock markets that later-stage startups rely on. While the FTT is pitched as a way to fund society in an AI-heavy financial era, critics see it as a blunt tool that could curb investment and innovation. They argue that if revenue is needed from the finance sector, it might be better to close loopholes or raise corporate taxes on financial firms' profits, rather than taxing each trade.

While FTT isn't framed explicitly as an "AI tax," it aligns with the goal of taxing automated trading that largely escapes income taxation. At congressional hearings, lawmakers have discussed adjusting depreciation and R&D tax rules that currently favor capital investment over labor – aiming to make the tax code neutral as to whether work is done by robots or humans. William Gale cautions that defining "robots" for taxation is problematic – if we tax "any labor-saving technology," then "your washing machine is a robot." This approach could hit some sectors arbitrarily. Is software automation a "robot"? What about an advanced machine that augments rather than replaces workers? We risk an overly broad or innovation-stifling policy. A broad-based tax capturing AI-driven wealth may prove superior to narrow robot taxation.

There are other tax proposals getting less attention than robot tax and FTT that merit discussion. A universal value-added tax (VAT), similar to what the EU has, taxes consumption at each production stage and ultimately passes to end consumers. When you buy a $100 product under a 10% VAT, you pay $10 in tax regardless of whether the product was made by humans or robots. VAT is collected incrementally at each stage of production or distribution. This could provide more stable revenue than relying on payroll and income taxes.

Then there is the digital services tax, which taxes the collection, processing, or monetization of data. France's digital services tax on tech companies is an example, though it's broader than just AI. Rather than directly taxing robots or AI, some propose fees based on the ratio of a company's revenue to its human workforce size. This creates incentives to maintain human employment while being more administratively feasible than trying to define and tax "robots." This "automation fee structure" could work alongside modified corporate taxes capturing more value from automated processes, such as limiting deductions for AI investments or creating new categories of taxable business income. Since AI may concentrate wealth among capital owners, some economists propose expanded wealth taxes to maintain revenue, including mark-to-market taxation of capital gains. Others suggest broadening what counts as "payroll," treating certain types of automated revenue as equivalent to wages for tax purposes.

Sam Altman, CEO of OpenAI, has proposed an "American Equity Fund" to ensure AI's benefits are widely shared. This fund would be capitalized by taxing companies above a certain valuation approximately 2.5% of their market value each year (in stock), as well as a 2.5% tax on high-value land, with proceeds distributed to citizens. This would make every citizen a shareholder in the nation's AI-powered growth. Creating a sovereign wealth fund that takes small ownership stakes in AI companies and distributes returns as a "tech dividend" to citizens is another approach. Alaska's Permanent Fund serves as a model.

Granted, Altman's proposal faces significant practical challenges. The administrative burden of assessing wealth for hundreds of millions of taxpayers would be enormous, and valuing non-market assets like private businesses and art presents particular difficulties. A more feasible approach might be to narrow the tax base to public securities and real estate, which have clearer market values. While these implementation hurdles are serious, they shouldn't overshadow Altman's core argument: Robot taxes would discourage AI innovation and drive companies to relocate overseas. This exodus would reduce U.S. tax revenue and create national security risks as competitor nations welcome displaced American companies.

Different countries will likely diverge in their approaches to AI regulation based on demographic challenges. Nations with aging populations, like Japan, may create AI-friendly policies to address their shrinking workforce. These countries might resist AI taxation and become attractive destinations for AI companies seeking lighter regulation. Countries facing high unemployment may implement more restrictive AI policies to protect existing jobs. This demographic-driven policy divide could create international tensions over AI governance, taxation, ethics, security, and copyright law.

Ideally, we should have global adoption of a unified corporate tax rate, eliminating incentives for companies to relocate for tax advantages. However, this vision is likely unrealistic. Countries will continue offering lower tax rates to attract overseas businesses.

We should also explore how AI can help reduce federal government spend, offsetting its impact on tax collection. During a House Oversight Committee hearing in 2023, a Deloitte study estimated that automated tasks performed by federal employees using AI could yield up to $41 billion in annual savings. The AI LEAD Act (H.R. 8756), introduced in the 118th Congress, directs federal agencies to explore how AI can improve government operations' efficiency.

A Congressional Research Service report highlighted that properly trained AI tools could improve clinical decision-making and patient outcomes in healthcare. In a 2023 Senate hearing, estimates suggested greater AI adoption could reduce national healthcare spending by 5% to 10%. If AI can meaningfully improve medical care, it opens doors for federal government to reduce Medicare spend, offering annual taxpayer savings.

Alternative solutions beyond taxation warrant consideration too. We must update the social contract for an AI-driven economy and ensure automation benefits are widely shared. Universal Basic Income (UBI) has been part of the discourse for years. Elon Musk suggests UBI will likely become necessary as AI displaces jobs, requiring redistribution of robot-generated wealth to meet basic needs. Others prefer job guarantee programs or wage subsidies, which may better incentivize continued work, arguing that UBI is fancier terminology for welfare. Many stress that the best "tax" on AI invests in people through reskilling, retraining, STEM education, apprenticeships, and job transition programs, using existing tax revenues to massively upscale workers. Some tech CEOs suggest tax incentives for companies that create jobs or retrain workers, rather than taxes on those that eliminate jobs, taking a carrot rather than stick approach.

So far, no AI-specific tax law has passed. Congress remains in early stages of these discussions, with privacy, bias, and national security vying for attention. DeepSeek dominates today's discussions, for instance. But the debate has begun on mechanisms – from robot taxes to closing loopholes – to offset impending tax base erosion.

Meanwhile, finance workers like Jen wait anxiously, wondering whether they will recover from post-AI extinction. As we navigate this transition, one thing is clear: whether through robot taxes, financial transaction fees, or creative solutions like equity funds, we must find ways to share in AI's prosperity without derailing technological progress that could benefit us all.