Thank you all for being here today at the WP studio. I'm Cat Zakrzewski and I'm a White House reporter here at The Washington Post. And my guest today is Michael Kratsios, the director of the Office of Science and Tech Policy at the White House. And he also has worked in the tech industry when he has not been in government, including for venture capitalists Peter Thiel and Scale AI. Michael, thanks so much for being with us here in Davos today.
Thanks so much for having me. It's great to be here.
And so we are here today, it's hard to believe because so much has happened, but on the one-year anniversary of Trump's inauguration. And I wanted to take a minute to give you a second to reflect on what have you learned about AI and government in this past year?
Well, this past year has been one of the most dramatic accelerations of the prioritization of technology in the history of American government. You know, I started my time working in government in Trump 45 and in the first term where I served as the CTO of the United States. And at that time, the president was interested in AI. He signed the first executive order ever in the history of the United States on AI in February of 2019. But technology was always kind of a tier 2 issue. It was something that was important, but it wasn't something that was on the front page of every newspaper. It was important for the economy, but it wasn't driving the stock market. It was important generally for relationships with foreign countries, but it wasn't the primary driver of a lot of these international deals that we're doing.
You fast forward only short four years and it is the story no matter what you're working on—if you're working on energy, if you're working on international relations, if you're working on national security. And this past year has sort of manifested itself across so many different portfolios. I think the key action, probably most important if you had to name one that the administration took in 2025, was all around the AI action plan. This was the plan the United States has to win the AI race that was released in July of last year. It lays out sort of three pillars of the strategy. One is about innovation: how does the US continue to out-innovate and make sure that the next great AI discoveries are made in the United States? The second is all about infrastructure and energy: how do we power the AI revolution that we need here in America? And the third pillar is all about international engagement: how do we work with our partners and allies to make sure that the whole world is running on an American stack? So I think to me, back to your question, it's about this prioritization of technology and its manifestation in so many different pieces of the policy puzzle.
And on that point, I want to come back to the AI action plan. But you mentioned just how this is so on the front page of newspapers, which is good for me. But I wanted to ask you a little bit about just the role the tech companies are playing, because it's quite different this time around. You saw a lot of the companies kind of keep the first Trump administration at arm's length, and now they're constantly in the Oval Office. How has that changed your job?
Well, I think companies are just one set of stakeholders that we have to deal with when we're trying to deliver results for the American people. And obviously, they have their interests and are doing an incredible job on developing new technologies, of commercializing those technologies, on exporting those technologies. You know, we've talked a lot in the past about research and development in the United States. The private sector now outspends the federal government on basic research and development by, you know, seven to three. So, 70% of basic R&D is by the private sector. So they play a very critical role in our larger innovation ecosystem. So I think it's natural that we have to hear from them as one of many constituents that we work with every day.
And that kind of relationship, though, has raised some concerns about conflicts of interest with this administration. What's your response to people who are worried about the closeness between some of the officials in the administration and the CEOs in industry?
Yeah, I'm not particularly familiar with that, but I think, as I said earlier, the key thing is we try to work with all constituents across the country, whether it be corporations, whether it be civil society, whether it be academia. And I think for us, having relationships with all those pieces of the puzzle allows us to make the best policy decisions for the country.
And recently there was an AI executive order around state AI laws that made a lot of headlines. And I wanted to ask you, part of that executive order directs your office and others to work on a national framework for AI policy with Congress. Can you give us a status update on that and what you expect next?
Yeah, that's a great question. I mean, zooming out a bit for those of you who aren't too familiar, I think this is born out of this larger policy question in the United States about whether or not a patchwork of state regulations around AI is a good or bad thing for the country. The president spoke about this particular issue when he gave his speech in July when we released the action plan, and the general position he took and the position of the administration is that creating a patchwork where there's 50 different rules around AI is actually a real challenge for innovators and borderline sort of anti-innovation. And I think the thing that ultimately happens with this patchwork is the people that suffer the most are usually the smallest technology companies or the ones that don't have armies of lawyers to find ways to navigate and make sure that they're complying with 50 different sets of rules. So the president directed me and David Sacks to work on a legislative proposal to create a sensible national framework for the United States. So that's what we're going to be doing in earnest for the next few months, and hopefully we'll have something to share in the near future.
And a lot of states, though, look at what's happening in Washington, and their argument for moving forward with their own laws is there's nothing on the books yet, and so we can act more quickly and we will. I mean, why should they wait for Washington to pass legislation given the gridlock we've seen in Congress?
Well, I think there's a couple things to think about. I don't necessarily buy the argument that there isn't anything on the books. A position that the Trump administration has taken since the first go-round in 45 is that when you regulate AI, you should think about it as a use case and sector-specific regulatory regime. And what we mean by that is having one individual AI law that covers everything related to AI is not feasible and it doesn't actually work. The Europeans attempted this with the EU AI Act, and it's been an abysmal failure. And the way that I like to think about it is that AI is going to be everywhere and it is everywhere. It's in medical diagnostics. It's in drones. It's in self-driving vehicles. It's in financial services. And the way that you regulate those is through the existing regulatory mechanisms that the federal government already has to regulate AI-powered medical diagnostics, to regulate drones, to regulate self-driving cars. And in this way, you actually have regulators that understand the industry that is being impacted by AI and create regulations and update regulations to a world that AI exists. So to me, I kind of push back against the fact that there aren't regulations—actually quite a bit. And if anything, we need to look at it to make sure we can, at some points, kind of accelerate places where AI can be deployed. But I think there is a balance to be struck, and the president was very clear in executive orders that there are places where the states should continue to look at regulations: things related to child safety, things related to the deployment of data centers and energy, and you wouldn't want to mess with that. So there are places certainly for the states to play.
And the order also directs the administration, though, to sue states over laws in certain other areas. And I just wanted to get a sense, you know, how do you square the need that you just mentioned to have national consistency with the rights of states to make their own rules?
Yeah, I think the answer is the same that I said. I think that the president made clear in the executive order that there are specific areas where states should continue to legislate, and we'd love to see that. And in places where we believe a sensible national framework makes sense, we'll hopefully draft something that can pass in Congress.
And I want to move on now to data centers, because this is an issue that we saw in some of the governor's races last year was really top of mind for voters who are really worried about rising energy costs in their own utility bills. And I just wanted to get a sense of what your administration is doing to ensure energy remains affordable while simultaneously pushing these data center buildouts.
Yeah. So I would say two things. I don't think you have to be sort of a politico that tracks the news to know that there's been no president more obsessed with energy abundance than President Trump. So everything that we possibly can do to drive energy production in the United States and affordability is being done, from permitting to licensing of different energy sources and so on. But I think the second piece, which is specific to the AI question, is around this idea of when a data center is being built, how do you ensure that the data center itself and the power that it uses doesn't ultimately result in ratepayers or regular Americans having increased electricity bills? And the truth is the president has been very clear on this fact. He actually did a Truth on this on Monday, which I'm sure you've seen, which he essentially said that if you are a data center, you should be paying your own way. You should pay for all the electricity that you will be consuming for your data center. You should pay for the production of that new electricity. You should pay for any new transmission lines that you are going to be creating in order to hook up that. And ultimately, I think as an administration, we're committed to ensuring that these data centers are not actually raising the rates of individual ratepayers. And layered on top of that is we've actually seen a lot of tech companies come out and make that pledge and make that promise. Microsoft just last week said that themselves, and I think many others are doing the same thing.
And on that point, though, can you trust these companies to make voluntary pledges just given the history that we've seen in this industry, or do we need laws on the books that would require that?
I don't want to speak for the companies, but I will say they are highly incentivized to do the right thing here. At the end of the day, the primary bottleneck to the buildout of these data centers in the United States is not federal laws or regulation; it is local zoning. And the challenge that they face, if you ask any of them today, the biggest challenge that these large data center builders face is local issues. It's whether or not they will get the permit at the local level to be able to build. So they need to make sure that they can convince these local communities that the rates aren't going to go up, or else they won't be able to actually build.
And on the topic of kind of voters' concerns about artificial intelligence, we see that job loss is a big fear for people. There was a Reuters poll last year that found 71% of respondents are concerned that AI will put too many people out of work permanently. What do you say to those people who are worried about the impact this is going to have on their jobs?
I would say that this administration is deeply committed to driving programs forward that can help retrain, reskill, and prepare America's workers for the economy of the future. There's large programs underway in the Secretary—Secretary of Labor has driven the Labor Department and Department of Education all for these reskilling programs and upskilling of Americans. So if transitions do happen, there are places where you can actually get the skills that you need to get a job that can work in the 21st century. I think the other thing that we think a lot about is how do we prepare young Americans for an AI future, and it's something that the White House has been very fixated on. The First Lady has been very active on this, and this is K through 12 education on AI and helping students in America understand where AI works really well, where it doesn't work, where you should use it, where you shouldn't use it. And the whole point is not necessarily to teach young Americans how to use this tool, but to make them understand the technology, to demystify it so that they can be ready for their job when they enter the workforce. And kind of the last piece, which I think is worth noting, is that what excites me about this AI future is there's going to be a whole set of jobs that are created which we have never heard of, haven't imagined, haven't thought of, and those are all jobs that are going to be filled by great American workers. And I'm excited for that future, and it's something that we as an economy have been able to transition through for many decades.
Going back to that reskilling program that you just mentioned with the Department of Labor, that was a key prong of the AI action plan. Can you tell us exactly what that program would look like and where the funding for it is coming from?
Yeah, I think the funding is coming from the Department of Labor. So that's an appropriated funding for reskilling that they're using. And these are programs, these are grant-related programs where folks that have reskilling opportunities can apply for grants that are then funded by the Department of Labor, that are then executed. And what's great about what the Department of Labor also does is they have strong partnerships with a lot of private sector companies that are also, they themselves, committed to this sort of retraining and reskilling initiative. So you can almost get sort of a 1 plus 1 equals 3 on that equation.
Are there any specific companies that are already doing this that you think show how this could work in the future?
