AI’s Upside: A Tax Plan for Progress

As society ponders the burgeoning AI revolution, it is quite reasonable to look at downsides of the Internet, Mobile and Social Media revolutions to consider what might have been done better. The 1990s era policy of not taxing the Internet may have helped fuel growth, but its legacy is seen in shuttered retail establishments once occupied by local businesses harmed by online tax arbitrage. Mobile wealth creation was captured by a very few companies which served to accelerate wealth concentration. Social media not only concentrated wealth but also unleashed societal forces that have led to isolation, depression and extremism. The emergence of AI gives the country an opportunity to draw learnings from these experiences and, perhaps, craft more just outcomes in the AI era.

With AI poised to broadly impact both GDP and employment, both economic and social policies must be top of mind. Fiddling with tax and investment policies is always politically hazardous. More practically, the law of unintended consequences has not been repealed, but complexity is no excuse for shrugging the country into laissez-faire AI policies. Keeping policies simple and the market in charge of picking winners/losers is important.

Successful AI linked economic policies must satisfy four requirements:

1 — All Americans must benefit

2 — Market forces must guide private sector investments

3 — Human skills must be prioritized

4 — Investment scale must align with real world AI deployment

Against these requirements, there are two places where the country should be willing to impose AI taxes in order to fund societal needs: Entitlement programs and education.

After generations of failure to address Social Security and Medicare funding shortfalls, the retirement crisis has arrived and the stakes are enormous. Within the next decade there must be significant entitlement reform in the form of reduced benefits or increased taxes. The explosion of AI growth — particularly as it is intertwined with both wealth concentration and human employment — creates the opportunity for society to fund these programs in a new way.

Education is another societal ill that will be deeply entwined with AI. College costs have soared, education debt has reached insane levels, and education is unattainable for many young people. At the same time, current workers whose jobs are displaced by AI must be armed with new skills for an AI world. It is sensible — indeed ethical — for society to require that some of AI’s economic gains be reinvested in education.

There are three places where taxes can be easily imposed to fund investments in these programs.

1–13.85% Tax on AI cloud services and AI Software Maintenance fees

Why 13.85%? Because that is the combined social security and medicare income tax rate for human labor. As AI technology takes up more labor, it must contribute back some of that income to provide needed social services. This easily clears all four of the criteria. Importantly, it leaves market forces in charge of the evolution and does not require a massive government intrusion/expense while the actual economic outcomes of AI remain unknown. This tax scales gracefully and removes tax arbitrage from AI vs. human business decisions. AI may be cheaper and more productive but not because companies capture the benefit of not paying for needed entitlement costs by displacing workers.

2 — Transaction Tax on Capitalized AI investments

A one time tax on AI investments should be put in place to help fund education. The tax should explicitly cover AI chip purchases and any capitalized software/labor or other AI related capital investments that are captured on company balance sheets. This will require a bit of accounting and enforcement but it is no more complicated than other accounting burdens the country places on corporations in order to properly manage depreciation schedules.

3 — $25k Education Credit for Displaced Workers

If AI results in large scale job displacements, the re-training burden on impacted individuals and society will be significant. Companies using AI to lay off workers should have to fund a $25K education credit specifically for each impacted individual to seek new skill training. A laid off call center worker who has a certificate to learn cybersecurity or prompt engineering skills will benefit greatly from those skills and society will reduce long term unemployment and its related ills.

Of course companies will endeavor to justify layoffs as not being AI related. Employees must have a vehicle to appeal that the layoff was AI related and companies found to be falsely claiming AI was not the cause must face substantial penalties.

Yes, collectively these moves will raise the bar on deploying AI — but not as much as it seems.

Businesses making AI productivity investments execute relatively simple business cases. Assume the total cost of a call center employee (salary, benefits, taxes, infrastructure) is $75,000/year. If it costs a company $25,000 to deploy AI that replaces an employee, then the company will get its money back in just four months. Burdening that business case with the above charges may add $30,000 one-time cost and perhaps a $1000/year recurring cost associated with the AI services. Under this approach the payback is perhaps extended to 10 months and delivers slightly less upside in the future. No CFO worth their salt will turn away an investment that provides perpetuity benefits and pays back within 10 months.

When combined, these tax policies would allow constructively link AI growth to needed societal investments. Importantly, these approaches do not pick winners or losers. Companies are not forced to maintain arbitrary employment levels nor are humans put at a disadvantage because they require social security funding. These approaches do not confer benefits on a specific technical architecture nor create arbitrage between in-sourcing and out-sourcing AI infrastructure. US chip and technology companies remain free to make sales outside of the US under whatever tax scheme other countries embrace.

No doubt there will be huge fights about any attempt to build programs off of AI taxes. The same bugaboos that were trotted out about internet taxes or other attempts to avoid wealth concentration will happen. There will be legitimate arguments about whether the government should use AI to address UBI (this is a bad idea) or go after climate change in some fashion. While laudable, those “boil the ocean” efforts are likely to stall due to sheer complexity.

Society has been presented with a very actionable moment to link AI to addressing specific social problems. Carpe Diem.

(Image Credit: Microsoft Copilot)