Articles of Interest: March 29 - April 4, 2025
- The Prometheus Security Project Team
- Apr 4
- 5 min read
Deborah Mary Sophia, Reuters. April 3, 2025.
President Trump's new tariffs could derail the U.S. tech sector's massive AI infrastructure investments. The steep duties of 34% on China, 32% on Taiwan, and a 10% baseline on all imports will significantly increase da ta center building costs.Even if semiconductors are currently exempt, other components needed for these data centers will still be impacted by the tariffs and are likely to drastically increase prices. This new economic policy threatens projects like the $500 billion Stargate venture, which Trump himself announced. Cloud providers like Microsoft, Alphabet, and Amazon already face investor skepticism over AI spending and these tariffs will likely lead to increased investor concerns as seen by the U.S. stock market's reaction to the Trump tariffs. They may now cut back further should they perceive these tariffs as detrimental to their investment in AI and AI infrastructure. Furthermore, tech stocks fell sharply after the announcement, as seen by Nvidia, AMD, and Broadcom losing 7-10% of their value.
Trump's Tariff plan emphasizes a confusing disconnect between the administration's economic strategy and national AI ambitions. While it is difficult to predict the im pact on data center costs, their cost will almost certainly rise and be felt by U.S. companies and investors. Companies like Microsoft and Google have been spending billions to build the massive computing centers needed for advanced AI. Now, they li kely will face a tough choice: spend much more than planned or scale back their ambitions. The U.S. risks hamstringing itself in the global AI arms race by increasing the domestic cost of AI infrastructure. Countries like China have grown accustomed to AI advancement under economic and supply chain restrictions, as seen by DeepSeek's success. China has shown that their AI innovation will not be easily curtailed by economic mechanisms, but can the same be said for U.S. AI companies?
Furthermore, beyond the U.S.-China AI competition, these tariffs are likely to have a negative impact on U.S. foreign policy with allies and adversaries. As AI technology becomes more advanced, the world will likely reach a point where international norms and governance are required to sufficiently address the global challenges. The U.S. and China will likely end up in opposite camps regarding global AI governance. Thus, coalition building will likely occur around the two global AI leaders. If the U.S. is at odds with its allies in Europe, Asia, and elsewhere, that coalition building becomes more difficult. These tariffs may push more countries to turn toward China for business and technology relations rather than the U.S.
Sophia, Deborah Mary. "Trump tariffs could stymie Big Tech's US data center spending spree." Reuters, April 3, 2025. https://www.reuters.com/technology/trump-tariffs-could-stymie-big-techs-us-data-center-spending-spree-2025-04-03/
Dan Hendrycks, The Economist. March 28th, 2025.
Dan Hendrycks argues against the U.S. pursuing a "Manhattan Project" approach to AI development. He believes that such an initiative would face insurmountable challenges including the impossibility of maintaining secrecy, the exclusion of vital international talent, and the destabilizing geopolitical consequences. Hendrycks warns that superintelligent AI presents unprecedented control risks and could upend nuclear deterrence. Instead, he recommends the U.S. focus on securing supply chains for AI infrastructure and developing capabilities to detect and disrupt competitors' superintelligence projects rather than focusing on our own.
While Hendrycks makes compelling points about the geopolitical tensions that would arise from an AI Manhattan Project, some degree of centralized coordination may still be necessary, and perhaps unavoidable. His concerns about China viewing such a project as an existential threat and responding with espionage and sabotage are well-founded and realistic, but similar tensions may emerge regardless of how AI advancement is structured, whether in the private sector or under government centralization.
Leaving superhuman AI development entirely to private companies creates its own geopolitical complications. Corporate interests may not align with national security priorities, potentially creating fragmented approaches to international AI governance and diplomatic tensions when companies operate across borders with different regulatory standards.
The alternative approaches that Hendrycks suggests (focusing on supply chain resilience and developing capabilities to disrupt rivals' AI projects) would likely still involve significant government coordination and could still trigger the same international escalation he warns against. Preemptive disruption of foreign AI development might actually accelerate an arms race rather than prevent one.
Hendrycks, Dan. "Dan Hendrycks warns America against launching a Manhattan Project for AI." The Economist, March 28th, 2025. https://www.economist.com/by-invitation/2025/03/28/dan-hendrycks-warns-america-against-launching-a-manhattan-project-for-ai?giftId=0630e2fb-f186-49f3-82d9-ffc429501c37&utm_campaign=gifted_article
Vlad Litnarovych, UNITED24 Media. April 2, 2025.
NATO has accepted multiple proposed systems from arms manufacturers throughout the West to use AI and drone swarms to stop Russian glide bomb attacks. Ukraine reportedly played a large role in determining the selected systems.
In the 15th Annual Innovation Challenge, NATO’s Allied Command Transformation department accepted 2 French and 1 German systems designed to tackle the threat of glide bomb attacks. Ukraine has repeatedly faced this threat throughout its war with Russia and has struggled to stop these attacks. As a result, several companies created a system of AI-powered drone swarms to address this problem.
The French company Alta created an AI-generated early warning system to track the trajectory of glide bombs. Germany’s Tytan Technology and France’s Atreyd created drone systems designed to intercept incoming glide bombs. This innovation is going to be a major step to stopping air attacks and yet another way AI has found itself being employed in a battlefield role. It also reveals that countries are identifying how drones can be considered a major defensive weapons system, not just used for air strikes and surveillance. The employment of drone swarms to stop air attacks can render some cities invulnerable to air attacks in the future.
Litnarovych, Vlad. “NATO Selects AI and Drone Swarm Solutions That Effectively Counter Russian Glide Bombs.” UNITED24 Media, April 2, 2025. https://united24media.com/latest-news/nato-selects-ai-and-drone-swarm-solutions-that-effectively-counter-russian-glide-bombs-7256.
Eleri Griffiths, BBC. April 3, 2025.
In 2020, Ben Black posted an April Fool's article saying that the Guinness Book of World Records recognized Cwmbran (a town in southern Wales) as having the most roundabouts per square kilometer. The story was then picked up by a larger news website, which published it without Mr. Black's permission and without verifying the claim.
This year, Google's AI tool presented the story as fact despite being labeled as an April Fool's joke. Mr. Black's article was not dangerous; it was intended to be a simple joke with the community. However, it highlights how easy it is for misinformation to spread rapidly and to be regarded as fact without any verification. Since the BBC article, Google AI has now highlighted the mistake and acknowledged the AI's impact on spreading misinformation.
AI tools are constantly improving, getting smarter, faster, and cheaper. Despite this, they are still susceptible to spreading false information. Without Mr. Black and the BBC talking about this, the 5-year-old prank may have continued to be considered a fact.
The story fooled both humans and AI alike, meaning we must remain skeptical and work to verify information independently whenever possible. This time, it was a joke about roundabouts, but next time, it could be something more serious.
Griffiths, Eleri. "Google AI presented my April Fools' story as real news." BBC, April 3, 2025. https://www.bbc.com/news/articles/cly12egqq5ko.
