The Dynamist

In 2025, tech policy felt like everything happened everywhere at once. Google lost two antitrust cases but avoided a breakup. Meta won its case entirely. The SEC went from suing crypto companies to dropping every major enforcement action. Net neutrality died—again—this time probably for good. TikTok got banned, then unbanned, then re-banned, then saved by executive order—five times. Chinese hackers compromised 200 companies through our telecom networks. And Congress finally actually passed a law protecting kids online—The Take it Down Act, to be precise.It was a year in many ways defined by tensions and contradictions. Courts stripped power from federal agencies just as the new administration tried to bring those agencies under tighter presidential control. The administration took some actions to be tough on China, while other measures appeared to let our chief adversary off the hook. States rushed to fill the vacuum on AI and privacy while the White House has threatened to preempt them. Platforms loosened content moderation in the US while facing record fines in Europe. And Washington declared it wanted to win the AI race—while local communities debated whether they even wanted data centers in their backyards.So what were the biggest tech and telecom policy stories of 2025? Which developments will have staying power, and which were little more than sound and fury? What should we be watching heading into 2026? And did anyone actually win this year—or did everyone just survive?To unpack all this, Evan is joined by Luke Hogg, Director of Technology Policy at FAI, and Josh Levine, Research Fellow at FAI.

China's October decision to add five rare earth elements to its export control list confirmed what policymakers have long feared. China controls 60% of global critical mineral production and over 80% of refining capacity for materials that power everything from electric vehicles to fighter jets. AI data center buildouts have only spiked demand further. Add cobalt to the picture—70% of global reserves sit in the Democratic Republic of Congo, and China owns roughly 70% of that production—and you have a supply chain built for peacetime that could collapse in a crisis. The alloys in today's F-35 engines depend on elements Beijing could cut off tomorrow.Joseph Krause argues the problem runs deeper than mining. Materials companies today are 75 to 150 years old. Some aerospace alloys still in use were developed for the Ford Model T. Meanwhile, China has been publishing the lion's share of advanced alloy research and aggressively recruiting metallurgy professors from American universities. China already fields a hypersonic capability using a niobium-based alloy; the US is scrambling to catch up. Krause's company, Radical AI, is building AI-powered labs to compress what typically takes 10 to 20 years and over $100 million in materials discovery into something dramatically faster and cheaper. The goal is inverse design: start with the exact properties the military needs, then work backward to find materials that don't require Chinese-controlled supply chains.The Trump administration has moved aggressively, taking a $400 million stake in MP Materials, putting $2 billion toward stockpiling strategic metals, and working to streamline permitting that currently takes seven to ten years for a single US mine. FAI’s Farrell Gregory notes there's no silver bullet across the 60 minerals on the USGS critical minerals list, which ranges from rare earths at $8 billion in global market value to copper at $250 billion. The administration has shifted from blanket tax credits to case-by-case deals, prioritizing materials where Chinese leverage is highest and American action can make the biggest difference. Krause and Gregory join Evan to discuss the challenges facing the U.S. amid Chinese dominance in rare earth minerals and what policymakers can do to make the U.S. more resilient to supply chain shocks, including public-private partnerships and government funding.

The federal government spends over $100 billion on information technology (IT) every year. About 80 percent of that goes toward operating and maintaining systems, many of which are long outdated and obsolete. Some federal IT systems are more than 50 years old.On day one of his presidency, Trump signed an EO that established the Department of Government Efficiency, which included a mandate to modernize “Federal technology and software to maximize governmental efficiency and productivity.”While DOGE helped shine a spotlight on the issue, it isn’t new. The Government Accountability Office has long warned about the risks of poor federal software practices—taxpayer waste, inefficient government processes, harms to citizens who rely on services like veterans benefits, and cybersecurity vulnerabilities.Many presidents have tried to solve it, but despite some improvements here and there, the problem has persisted for lots of reasons. Government agencies often lack the expertise to understand their software products and needs. Agencies have also failed to properly audit and track their software purchases. The companies who sell software to the government often deliberately make it difficult for agencies to modernize, change vendors, or diversify their supply chains.With a renewed focus on government efficiency, how can Congress and the Trump administration tackle the long-festering problem of outdated and vulnerable federal IT? What can agencies do on their own, and what requires an act of Congress? And how would the American people benefit from improving these systems?Evan is joined by Dan Lips, Senior Fellow at FAI and Luke Hogg, Director of Tech Policy at FAI. For more, see Dan’s blog post and Evan’s op-ed. 

The push for a federal standard on AI is back. With support from President Trump, House Majority Leader Steve Scalise is looking to add an effective ban on state-level AI regulation to the end of year National Defense Authorization Act. Despite the White House’s backing and strong support from the tech  industry, the effort is facing bipartisan pushback, including from Republican governors like Florida’s Ron DeSantis and Democrats in Congress.The battle is shaping up to be a redux of the moratorium effort from the summer, when a ban on state AI rules came close, but failed to make it into the One Big Beautiful Bill. While that preemption effort didn’t come with any federal standards in its place, this time proponents of federal preemption are working to assure skeptics that this won’t just be a ban on state rules, but will establish some federal safeguards on AI safety and child protection.Can Congress agree to create a national standard that goes beyond simply telling states what they can’t do? Have the politics changed much since July when the prior effort failed? Will proposed safeguards be enough to move skeptics and those concerned about AI’s societal impact?Evan is joined by Dean Ball, senior fellow at FAI. Previously, he was Senior Policy Advisor for Artificial Intelligence and Emerging Technology at the White House Office of Science and Technology Policy and the primary staff drafter of America’s AI Action Plan. He is the author of the Hyperdimensional Substack, where his work focuses on emerging technologies and the future of governance.

The future of AI may be decided in backyards. Data Centers—the sprawling facilities designed to support the massive computing required to train and run AI models—are being built across the country. One estimate sees more than $1 trillion dollars in capital spending on data centers in the next four years. And they use electricity—a lot of it. While data centers can bring construction jobs,  tax revenue, and economic development to their communities, they also bring complaints about power and water usage, noise pollution, and architectural blight.Debates are raging from town halls to the halls of Congress. Yes, politicians want the US to lead the world in AI, but elected officials, particularly local ones, are hearing from constituents concerned about data centers, including the potential to raise electric bills. The decisions being made right now in places like Northern Virginia, Umatilla, Oregon, and Mount Pleasant, Wisconsin, will determine whether AI infrastructure is scaled quickly, or whether a backlash slows it down. If done right, data centers can bring world-class tech capabilities, lower electricity prices, energy abundance, and local tax revenue. Done poorly, we see working class Americans paying more for power, the electric grid struggling, and the potential for the American public to turn sour on data canters en masse.So what do people need to know about data centers to make informed decisions? What really is the impact of data centers on water and electricity? What should policymakers in Washington do, if anything, about these debates? And are there ways to balance legitimate local concerns without hamstringing a strategic imperative?Evan is joined by Asad Ramzanali, Director of Artificial Intelligence & Technology Policy at the Vanderbilt Policy Accelerator. He was previously Chief of Staff at the White House Office of Science and Tech Policy under President Biden and Legislative Director to former Rep. Anna Eshoo (D-CA). You can read his recent op-ed on data centers here. Evan is also joined by Daniel King, Research Fellow at FAI where he focuses on the energy and security dimensions of artificial intelligence. Daniel completed Master's studies in Statistics & Data Science at Yale University and earned a B.S. in Mechanical Engineering and Applied Mathematics from Brown University. Check out his substack on AI and energy, Policy Gradients.