Skip to content
Should we tax AI? episode artwork

Jun 10, 2026 9 min

Should we tax AI?

from The Indicator from Planet Money

Should we tax AI? The AI race has made a lot of people richer … but most of those gains seem to have gone to the wealthy while everyday workers aren't seeing their incomes go up. We speak with Democratic congressional candidate, Alex Bores, who wants to tax AI. Also, we talk to a tax expert who has her reservations. Fact checking by Tyler Jones. Your Next Listen — It's come to this: Human certification in the age of AI slop Connect with The Indicator — Sign up for The Indicator’s brand new newsletter — Buy the Planet Money book — Find our socials, YouTube and more! — For sponsor-free episodes, subscribe to NPR+ NPR Privacy Policy

Episode details come from the podcast's RSS feed.

Embed this episode

Ready to play

Should we tax AI?

0:00 9:10
of matches

Transcript Auto-generated, may contain errors

NPR. There's a lot of fear out there that AI could eliminate a ton of jobs, form a permanent underclass, and enrich just a few people. Right now, we haven't seen large layoffs, but tech hiring is slow, and we've seen huge growth in wealth at the top since the launch of ChatGPT. Yeah, just scan the rich list. Elon Musk, Oracle's Larry Ellison, Mark Zuckerberg, Jeff Bezos, Google's Larry Page. These are all people whose wealth is ballooning with the race to build AI. The richest 26 Americans' wealth has grown an estimated 127% since November 2022.

And so if this is just a taste of the future, some lawmakers are demanding we tax AI and give the rest of us an AI dividend. This is The Indicator from Planet Money. I'm Darian Woods. And I'm Waylon Wong. Today on the show, should we tax AI? The AI race has made a lot of people richer, but most of those gains seem to have gone to those already with metaphorical bathtubs full of money. Yes. Meanwhile, everyday workers aren't seeing big pay raises. We speak to a congressional candidate who wants to tax AI, and we talk to a tax expert who has her reservations.

This week on the NPR Politics Podcast, the midterm elections are wrapping up and millions of people have already voted. We're taking one final check of the electoral pulse with a new NPR PBS News Marist poll. It shows Democrats still poised to make gains, but also with some warning signs for just how many. We'll explain on the NPR Politics Podcast. The countdown to the 2026 midterm elections is here. Voters will determine who controls the Senate, the House, and countless local offices.

All the election coverage you need to be found on the NPR app, from big picture political headlines to local stories that are curated just for you. Plus, live coverage as results come in on election night. Download the NPR app at npr.org slash 2026 election. This message comes from NPR sponsor, the NPR Wine Club, a place to find boutique wines from around the world and bottles inspired by popular NPR shows, as well as to explore the producers, regions, and exciting stories that bring each selection to life. Whether you buy a few bottles or join the club, all purchases support NPR programming and help fund quality reporting. You can learn more at nprwineclub.org. Must be 21 or older to purchase.

New York State Representative and current Democratic congressional candidate Alex Boras has a range of measures he's proposing to fund an AI dividend. Alex Boras is an interesting politician on AI because he really knows the technology. He's a computer scientist who's actually worked at machine learning startups. I'm saying what all Americans are saying, which is this is happening way too quickly and government right now doesn't seem up to the task to protect us. Alex's AI dividend would be a direct cash payment to Americans. It would also pay for jobs retraining and education to help workers laid off by AI.

Alex has two main ways he's proposing to tax AI, which will help fund this dividend. The first is by reducing the ability for businesses to deduct AI costs as a normal business expense. Right now, our tax structure actually incentivizes laying people off to invest in AI. We put huge taxes on hiring a human or keeping a human hired. We have employment taxes, we have payroll taxes, we have requirements for benefits, and we put huge discounts on using AI.

At the moment, AI is generally not taxed. In some states, like here in New York, I'll pay a sales tax if I use AI. But for businesses using AI, let's say they use ChatGPT, they'll actually deduct that expense from their revenue, so they pay less tax. Alex's second proposal to raise government revenue from AI is what's called a token tax. So a token is the basic unit of these large language models, and it can be either a small word or a piece of a word, but it is the computational unit that a model uses in processing language and understanding.

So asking a chatbot for, I don't know, a lunch recommendation might only use a few tokens. But asking it to generate an essay on the major themes of War and Peace would use a lot more tokens. Large AI users are charged by the token. As of this recording, OpenAI charges a standard rate of up to $30 for 1 million tokens of output. So that War and Peace essay, that'll cost you a few cents. OpenAI, what is it good for? Alex says we can think of this token tax as just a tax for commercial AI use. Subscriptions, data usage, whatever. The basics of it are taxing the commercial use of AI.

times when you're using AI to replace a human. But taxing AI might have unintended consequences. Martha Gimbel is the executive director of the Budget Lab, a policy think tank at Yale University. The thing that I think is a little bit frustrating for the tax nerds, so, you know, for the 15 of us, is why, right? Like, what are you trying to do? Martha says the reason that the why is important is because addressing different problems means using different policies.

So if you are interested in decreasing energy usage, you would design the tax one way. If you are trying to discourage the use of AI overall, then you might design something like a token tax. Martha says she's skeptical of a token tax because it's a tax on businesses' inputs. So for a call center, your inputs might be office supplies, rent, workers, and an AI subscription to help answer customers' questions. In general, economists tend to like taxing the final output rather than inputs because it means that you don't produce as efficiently. This is the same reason that economists don't like...

Like steel and aluminum tariffs. Yeah, it's like the government steering how businesses should be run, not leaving that to business managers. Martha says what would happen if it turned out that not that much AI computing power was required to replace all the call center workers? Yeah, I mean, like we all like to think that our problems are individual and special. Well, mine are. But most of us calling this call center, we have the same issues. And maybe the AI bot doesn't have to work very hard. They don't need that many tokens. And the token tax isn't aimed at...

getting at the increasing profit that that company is seeing from that because there isn't as much AI usage as there was relative to the number of workers that they had. Martha says, contrast this with a hypothetical biochemist who might want to use a lot of AI tokens. If you're taxing per token and you have a scientist in a lab trying to cure cancer and they're using a ton of tokens, they then bear the cost of the token tax. Whereas the call center that doesn't require as many tokens to replace its workers. doesn't. Martha would prefer that lawmakers look at existing policy gaps that would be an issue with or without a big wave of job losses due to AI. A standard corporate tax or reform in the capital gains system, things like that.

will do a better job of capturing the revenue that's generated from that. In other words, Martha says, focus on the loopholes in the current system that means companies in general don't pay their fair share of taxes from profits. She also wants the unemployment insurance system to be more generous to help laid off workers, whatever the cause. We brought these critiques to lawmaker Alex Boras. Are there existing tools that are more about general inequality or general ways that we tax corporations that don't kind of presuppose a particular pathway for the technology to evolve? Why does AI need to be taxed specifically?

AI is the first technology ever developed where the makers of the technology are explicitly saying that they're trying to replace all human labor. Every other technology before was by default a complement to human beings, not a substitute. Now, it doesn't mean that the labs will succeed, but having a tax to deal with that difference, that this very well could be about replacing human beings, is narrowly tailored and targeted at exactly the thing we're trying to get done. Are you trying to slow down the adoption of AI? And if so, do you think there's a risk that American rivals like China might advance AI faster if the U.S. is taxing AI and they're not?

I don't primarily think so, but I also think that we shouldn't be incentivizing replacing humans with AI. That's the current policy. I think we should continue research into AI, especially AI safety. But if we're constantly thinking about the race to roll it out everywhere and to everyone, we're going to devastate the economy. We're going to devastate the job market. We're going to devastate our kids. That's not the way to win this race with all of those costs. Right now, AI hasn't devastated the labor market. Martha points to a census survey of businesses. It shows that 96% of them haven't changed their hiring because of AI, and 2% have actually increased their hiring because of AI.

But if that starts to change, you can bet that policymakers will be looking at proposals like Alex's. The 2026 midterm elections are upon us. On November 3rd, voters will determine what the next two years of their local leadership and the balance of national government will look like. Listen to the NPR News Now podcast for all the updates you need ahead of and after November 3rd from NPR News.

This midterm election season, the NPR app is your single stop for clear, independent coverage. Stay on top of national and local storylines with news, stories, podcasts, and political analysis on demand. Download the NPR app and check out the Elections 2026 tab today.

Frequently asked questions

How long is this episode of The Indicator from Planet Money?

This episode is 9 minutes long.

When was this The Indicator from Planet Money episode published?

This episode was published on June 10, 2026.

Is there a transcript available for this episode?

Yes, a full transcript is available for this episode. You can read the complete transcript on the episode page.

Can I download this The Indicator from Planet Money episode?

Yes. Use the download control on the episode player to save the publisher-provided media file.