Delivered to your inbox every Monday, Last Week in AI Policy is a rundown of the previous week’s happenings in AI policy and governance in America. News, articles, opinion pieces, and more, to bring you up to speed for the week ahead.
On April 2, on his self-styled “liberation day", President Trump put in place sweeping new tariffs on countries around the world. A long-time advocate of tariffs, the President’s actions raise tariff rates to their highest in over 100 years. This move has sent markets into turmoil and placed many industries at risk. Advocates of the plan suggest, variously, that it will: bring in trillions in revenue, force countries to the negotiating table to then eliminate their tariffs and usher in a new era of free trade, or reshore American manufacturing.
Leaving aside that these goals directly contradict one another, the market doesn’t seem to believe that any of them will come to pass (without damaging the long-term profitability of American business, at least). While notionally directed at manufacturing, the tariffs have already wiped $1 trillion off the market capitalization of U.S. tech businesses. Crucially, by placing tariffs on critical components like GPUs, they may raise the price of developing leading AI models in the U.S., just as China appears to be closing the gap.
Federal Government
Tariffs Threaten GPU Supply Chains
New tariffs would impose a 32% rate on fully assembled GPU servers, primarily imported from Taiwan, which are integral components in data centers and AI research facilities nationwide.
This policy change has resulted in significant market instability, with tech giants including Nvidia, Amazon, Google, and Microsoft collectively losing over $1 trillion in market capitalization following the announcement.
Nvidia responded by planning a shift in manufacturing operations to U.S. facilities in an attempt to mitigate long-term tariff impacts.
Industry stakeholders expressed concern over the ambiguity of the tariff regulations, noting a lack of clarity regarding whether semiconductor chips themselves, integral to GPUs, are fully exempt.
"There's no doubt that the equipment that goes into data centers will become significantly more expensive," noted D.A. Davidson analyst Gil Luria, forecasting delays and possible scaling back of major AI infrastructure projects such as Stargate, a $500 billion initiative backed by OpenAI, Oracle, and SoftBank.
Were Tariff Rates Calculated Using ChatGPT?
Economists raised questions about the methodology used by the Trump administration, noting that several LLMs produced remarkably similar approaches when asked to design a method for calculating tariff rates.
Economist James Surowiecki further questioned the rigor of the Trump administration’s approach, calling the tariff calculations "extraordinary nonsense.”
CREATE AI Act Introduced to Democratize AI Research
Representatives Jay Obernolte (R-CA) and Don Beyer (D-VA) introduced bipartisan legislation to create the National Artificial Intelligence Research Resource (NAIRR), aiming to democratize access to vital AI research infrastructure across America.
The CREATE AI Act aims to provide researchers, educators, startups, and small businesses with equitable access to compute, curated datasets, and educational tools.
Chinese Firms Place $16 Billion Order for Nvidia Chips
Major Chinese technology companies, including ByteDance, Alibaba, and Tencent, collectively placed orders totaling $16 billion for Nvidia's H20 in Q1 of 2025.
While this is seemingly good news for Nvidia in the immediate term, some suggest they face a dilemma: capitalize on increased demand now, and become dependent on supply chains at risk of export restrictions in the future.
OMB Issues Updated AI Governance Guidelines
The Office of Management and Budget (OMB) released updated guidance on the adoption and management of artificial intelligence across federal agencies.
The guidelines promote data sharing and refined procurement practices, in a effort to accelerate AI use in domestic government.
State Government
New Jersey Criminalizes Deceptive AI-generated Media
Governor Phil Murphy signed legislation that would criminalize the creation and dissemination of deceptive AI-generated content, commonly known as deepfakes.
Penalties include up to five years imprisonment, alongside provisions allowing victims to pursue civil actions.
Activists, including deepfake victim Francesca Mani, helped shape the legislation, which forms part of a broader attempt by a growing number of states to crack down on the dissemination of deepfakes.
Press Clips
The Early Death of International AI Governance (Threading the Needle)
The Building Blocks of an AI Axis? (Special Competitive Studies Project) 📽
Understanding U.S. Allies’ Current Legal Authority to Implement AI and Semiconductor Export Controls (Centre for Strategic & International Studies)
The State and Local AI Regulation Landscape with Dean Ball (The AI Policy Podcast) 📽
Is China Racing to AGI? (ChinaTalk)
What we learned from reading ~100 AI safety evaluations (AI Policy Perspectives)
AI 2027: What Superintelligence Looks Like (Daniel Kokotajlo et al.)
2027 Intelligence Explosion: Month-by-Month Model (Scott Alexander & Daniel Kokotajlo) 📽
Is it 3 Years, or 3 Decades Away? (Epoch AI) 📽
How state AI regulations threaten innovation, free speech, and knowledge creation (The Eternally Radical Idea)
Where We Are Headed (Part II) (Hyperdimensional)
An Approach to Technical AGI Safety and Security (Google, Deepmind)