Artificial intelligence will play an increasingly important role in cyber defense, but vulnerabilities in AI systems call into question their reliability in the face of evolving offensive campaigns. Because securing AI systems can require trade-offs based on the types of threats, defenders are often caught in a constant balancing act. This report explores the challenges in AI security and their implications for deploying AI-enabled cyber defenses at scale.

Artificial intelligence offers enormous promise to advance progress, and powerful capabilities to disrupt it. This policy brief is the first installment of a series that examines how advances in AI could be exploited to enhance operations that automate disinformation. Introducing the RICHDATA framework—a disinformation kill chain—this report describes the stages and techniques used by human operators to build disinformation campaigns.

Stanford HAI Director of Policy Russell Wald, CSET Senior Fellow Andrew Lohn and Stanford HAI Postdoctoral Fellow Jeff Ding discussed how a National Research Cloud will impact U.S. national security.

In science and technology, U.S. federal prize competitions are a way to promote innovation, advance knowledge, and solicit technological solutions to problems. In this report, the authors identify the unique advantages of such competitions over traditional R&D processes, and how these advantages might benefit artificial intelligence research.

Created through a joint partnership between CSET and the AI Education Project, the AI Education Catalog aims to raise awareness of the AI-related programs available to students and educators, as well as to help inform AI education and workforce policy.

CSET submitted this comment to the Office of Science and Technology Policy and the National Science Foundation to support the work of the National Artificial Intelligence Research Resource (NAIRR) Task Force to develop an implementation roadmap that would provide AI researchers and students across scientific disciplines access to computational resources, high-quality data, educational tools, and user support.

Software vulnerability discovery, patching, and exploitation—collectively known as the vulnerability lifecycle—is time consuming and labor intensive. Automating the process could significantly improve software security and offensive hacking. The Defense Advanced Research Projects Agency’s Cyber Grand Challenge supported teams of researchers from 2014 to 2016 that worked to create these tools. China took notice. In 2017, China hosted its first Robot Hacking Game, seeking to automate the software vulnerability lifecycle. Since then, China has hosted seven such competitions and the People’s Liberation Army has increased its role in hosting the games.

The United States and China are keeping an eye on Indonesia’s artificial intelligence potential given the country’s innovation-driven national strategy and flourishing AI industry. China views Indonesia as an anchor for its economic, digital, and political inroads in Southeast Asia and has invested aggressively in new partnerships. The United States, with robust political and economic relations rooted in shared democratic ideals, has an opportunity to leverage its comparative advantages and tap into Indonesia’s AI potential through high-level agreements.

CSET research staff discussed the potential for OpenAI's GPT-3 and other content generation systems to be used as tools for turbocharging disinformation.

To what extent does China’s cultivation of talent in cybersecurity and AI matter in terms of competitiveness with other countries? Right now, it seems to have an edge: China’s 11 World-Class Cybersecurity Schools offer more classes on artificial intelligence and machine learning than do the 20 U.S. universities certified as Centers of Academic Excellence in Cyber Operations. This policy brief recommends tracking 13 research grants from the National Science Foundation that attempt to integrate AI into cybersecurity curricula.