Working Papers

The Risks of Generative AI for Grey Zone Conflict: Information Operations

The rise of generative AI has raised serious concerns about its impact on the information environment, particularly its ability to enhance the reach, precision, and sophistication of information operations. By lowering the barriers to entry and enhancing the persuasiveness of propaganda, generative AI can enable malign actors to conduct scalable disinformation campaigns that undermine traditional methods of attribution and response. While the potential for deception is substantial, we evaluate these concerns through evidence drawn from contemporary cases of generative AI in grey zone conflict. Ultimately, the chapter argues for a more measured view of AI’s role in information warfare, one that recognizes both its capabilities and its limitations within the broader context of grey zone conflict.

*This is a co-authored paper on which I am the first author

Influence Beyond Content: Generative AI and the Infrastructures of Information Operations

Generative AI has become a central concern in debates about contemporary information operations, particularly because of its potential to accelerate the production of deceptive content. This focus, however, misses some of the other ways generative AI is being used to enhance the underlying infrastructures of manipulation. Drawing on private-sector threat reporting and documented cases of Russian and Chinese information operations since 2022, this study examines the integration of generative AI across the information operations kill chain, including asset acquisition and disguise, information gathering and planning, and engagement. We find that generative AI primarily functions as an operational accelerant across multiple stages of information operations. These developments raise new challenges for detection, attribution, and governance, with implications that extend beyond platform moderation to AI system oversight.

*This is a co-authored paper on which I am the first author

Ongoing Projects

Are LLMs Transforming Cybercriminal Operations? A Study of LLMs on Dark Web Marketplaces

This project, funded by the AI Security Institute (AISI), United Kingdom Department of Science, Innovation and Technology, analyzes data from dark web forums, where a recurring theme is the sharing of "jailbreaks" — prompts designed to bypass commercial LLM safeguards and repurpose models for unintended uses. We argue that these data can 1) provide a critical test of existing jailbreak research and 2) provide information on how jailbreaking is discussed in these forums and how effective the shared prompts are. The findings from this study have implications for assessing the external validity of audit-based AI safety research and for understanding how model safeguards perform under real-world adversarial conditions.

*This is a co-authored project

Governance by Contract: AI Procurement as a Pathway to Democratic Erosion

Federal agencies frequently acquire and utilize AI-enabled surveillance systems – facial recognition platforms, OSINT tools, autonomous monitoring programs – through procurement contracts. Procurement has become a systematic institutional pathway through which governance authority migrates toward executive agencies and private vendors, and that this, in turn, contributes to democratic backsliding. This paper contains three arguments: 1) AI surveillance systems are almost universally acquired through executive procurement rather than through legislative authorization; 2) vendors’ proprietary technology claims and the technology’s complexity produce an opacity that prevents scrutiny; and 3) vendor relationships are self-reinforcing. Drawing on an original dataset of 145 federal contracts, this paper identifies procurement-driven authority reallocation as a form of de facto executive aggrandizement that operates below the threshold of political visibility and resists the accountability mechanisms democratic systems rely on to detect and counter executive overreach.

Publications

AI Governance and the Geopolitics of Extraction

The infrastructure of artificial intelligence (AI) depends on critical minerals and rare earth elements, making their supply chains a central factor in national security, economic stability, and global technology competition and governance. Overreliance on supply chains characterized by limited geographic diversification exposes AI commercial and defence industries to trade disruptions, cyber sabotage and strategic leverage. Diversifying supply chains requires international cooperation that accounts for sustainability, labour standards and transparency to balance national security with environmental and human security concerns.

You can read the full policy brief at this link.

Iranian Influence Operations

The state of Iran, directly and through proxies, has become increasingly active and sophisticated in online activities targeting the U.S. government and the American public. As the U.S. takes steps to enhance the security of its election infrastructure, promote public confidence in the democratic process, and combat disinformation and misinformation, the Iranian threat merits close attention.

You can read the full whitepaper at this link.