JD Vance speaks at the AI Action Summit in Paris. | Nathan Laine/Bloomberg via Getty ImagesOne of the big challenges of reporting on the new administration is that it’s hard to figure out who’s really calling the shots. Take AI. In his comments to world leaders at the AI Action Summit in Paris last week, Vice President JD Vance laid out one view of where the Trump administration should go on AI: “This administration will ensure that American AI technology continues to be the gold standard worldwide and we are the partner of choice for others — foreign countries and certainly businesses — as they expand their own use of AI.” He specifically spoke about the importance of America having its own advanced semiconductor design industry. Right now, there’s a desperate scramble in the AI industry to source enough of the advanced computer chips on which AI systems run. They’re an essential technology, and building them in America is a reasonable priority whether you’re concerned about the capabilities of advanced AI, conventional war with China that would dampen semiconductor imports, or reshoring manufacturing. Vance’s speech was bullish on AI: Based on his remarks, I expected that while many features of Biden administration-era AI policy, like concerns about AI-generated misinformation, were probably on the outs, the 2022 CHIPS and Science Act — which encouraged the development of an advanced American semiconductor industry, and restricted exports of advanced semiconductors — was here to stay. And so far the CHIPS project has been going quite well. So I expected the administration to stick with it or even double down.This story was first featured in the Future Perfect newsletter.Sign up here to explore the big, complicated problems the world faces and the most efficient ways to solve them. Sent twice a week.Trump officials might still do so. But it’s now hard to tell, because of the actions of a different arm of the new administration: DOGE. As you’ve surely heard by now, Elon Musk’s so-called Department of Government Efficiency is across many government departments systematically firing all probationary employees — that is, broadly anyone who was hired or promoted in the last year. That includes most of the people in the government who have been hired to work on AI, including a majority of the staff working on implementing the CHIPS Act, Axios reported this week, citing conversations with sources familiar with the decision. Per Axios, “74 postdocs, 57% of CHIPS staff focused on incentives,” and “67% of CHIPS staff focused on R&D” were slated to be fired. (It appears the cuts haven’t happened yet, so it’s still very possible the admin will reverse course or limit their scope.)To be clear: This isn’t a targeted cut. It’s not that the administration has decided to reverse course on CHIPS. It’s just that, in the course of firing every single recent hire across broad swathes of the government, Musk’s team may by accident fire all of the recent expertise that has been accumulated around AI.AI moves fastAt the start of the Biden administration, almost no one outside Silicon Valley was talking about AI. Even by the 2022 midterm elections, we were still pre-ChatGPT. It is only in the past two years that AI has begun to turn the world upside down. First there was ChatGPT, and then the scramble from every other major tech company in the US to produce a competitor: DALL-E 2 and then Midjourney, and a flowering of other open-source AI art generators, drove artists out of business and transformed the visual content available on the internet. Then came DeepSeek and other AI models that can “reason,” and now Deep Research, which many people have said can essentially replace research interns. Because it has all happened in the space of two years, all of the government hires focused on grappling with our new AI-powered world are new hires — including not just the ones working on CHIPS, but also everybody at the National Institute of Standards and Technology’s AI Safety Institute, which does AI research and tests the risks and capabilities of commercial AI models. And that means most of them are probationary hires who are likely to be impacted by DOGE’s rampage through the federal government.This is an extremely stupid problem to have. If the administration wants to give up on an advanced semiconductors industry in the US, that’s one thing. But if the administration wants to have an advanced semiconductors industry to remain competitive on the most important technology of the 21st century, as Vance said it does last week, it is absurd to give it up because it so happens that we hired most of the people who work on it quite recently. One of the ways that AI poses an unprecedented challenge for the government is just how fast it moves. It’s hardly the first technology to obviate lots of jobs, or to dramatically change the way other jobs work. But the faster those transitions happen, the more lives they can damage in their wake, and the harder it is for democratic mechanisms of oversight and regulation to function. If AI were going to happen slowly, gradually changing the way we live and work over the next 30 years, I’d be fairly optimistic that we’d have plenty of time to course correct, learn, and design appropriate laws. But AI is happening very fast, and the US government — which is not generally very good at doing things fast — has been continually on the back foot. And DOGE looks liable to make that a lot worse — not even through malice, but through inattention. Firing people is a lot faster than hiring them; freezing funding is a lot faster than figuring out where it can be put to best use. But with AI, we need to be able to accumulate expertise quickly, act on it quickly, and stay abreast of the pace of change in the private sector. DOGE was originally conceived as an agency that would make that easier. But so far, they’re making it harder. It seems very likely to me that world-transforming AI will arrive during the current administration, technology that may replace tens of millions of jobs overnight, that will represent a massive change in concentration of power in our society, and that experts warn could go fatally wrong. Vance isn’t wrong about the upsides. Musk, who believes there’s a very real chance that AI will destroy humanity, isn’t wrong about the downsides. But they’re both going to have to get serious if they want their administration in a position to make this go well.