On October 24, the White House issued the first-ever National Security Memorandum on Artificial Intelligence. CSET's experts answer pressing questions and what it means for U.S. national security and AI development.

This blog describes key takeaways from the NATO-Ukraine Defense Innovators Forum, held in Krakow, Poland in June 2024. It overviews changing concepts of operation, battlefield realities, and technological aspirations and innovations in Ukraine, with a focus on uncrewed aerial vehicles (UAVs) and counter-UAV systems. It builds upon CSET’s previous blog from the Future of Drones in Ukraine conference held in Warsaw in November 2023.

This year, CSET researchers returned to the DEF CON cybersecurity conference to explore how understandings of AI red-teaming practices have evolved among cybersecurity practitioners and AI experts. This blog post, a companion to "How I Won DEF CON’s Generative AI Red-Teaming Challenge", summarizes our takeaways and concludes with a list of outstanding research questions regarding AI red-teaming, some of which CSET hopes to address in future work.

In August 2024, CSET Research Fellow Colin Shea-Blymyer attended DEF CON, the world’s largest hacking convention to break powerful artificial intelligence systems. He participated in the AI red-teaming challenge, and won. This blog post details his experiences with the challenge, what it took to win the grand prize, and what he learned about the state of AI testing.

Artificial intelligence is becoming more integrated into the sciences. One of the scientific fields experiencing this is computational biology, which uses computer modeling to understand biological mechanisms and systems. This blog post provides an understanding of important research trends in these subject areas, and how advancements in AI can improve the speed and efficiency of computational biology to improve human health and well-being.

This blog post recounts the development of a Python-based data collection pipeline project completed in the summer of 2024 by CSET inaugural intern Jordan Monts. During his project research and work, he used the Requests and BeautifulSoup libraries to create a two-part system to gather and process web data, support ongoing research initiatives, and strengthen his skills in data processing as well as application programming interface (API) management.

This blog post analyzes five major Chinese AI governance proposals, focusing on the key actors specified in each proposal. We find that older proposals lack specificity when identifying AI governance actors. Recent proposals, on the other hand, assign roles and responsibilities to a defined set of actors. The findings from this blog post can help policymakers and analysts better understand China’s fast-evolving AI governance landscape.

Since 2022, U.S. export controls have restricted the highest-performing AI chips from being exported to China. The Biden administration likely did not intend to control CPUs (i.e., general-purpose processors) with these restrictions. However, CPUs are increasingly subject to export controls because chip designers are incorporating specialized elements for AI computation into CPUs. In this blog post, we discuss the implications of controlling AI-capable CPUs and make recommendations for the Bureau of Industry and Security (BIS) at the U.S. Department of Commerce.

The European Union's Artificial Intelligence Act has officially come into force today after more than five years of legislative processes and negotiations. While marking a significant milestone, it also initiates a prolonged phase of implementation, refinement, and enforcement. This blog post outlines key aspects of the regulation, such as rules for general-purpose AI and governance structures, and provides insights into its timeline and future expectations.

Researchers, companies, and policymakers have dedicated increasing attention to evaluating large language models (LLMs). This explainer covers why researchers are interested in evaluations, as well as some common evaluations and associated challenges. While evaluations can be helpful for monitoring progress, assessing risk, and determining whether to use a model for a specific purpose, they are still at a very early stage.