Overseas malign information operations by foreign actors seek to undermine the strategic interests of the United States. These operations are intended to manipulate the global information environment for geostrategic purposes by disseminating false or misleading information to shape narratives, shift public discourse, and undermine other nations' national security. These operations often leverage classical propaganda techniques, such as name-calling, glittering generalities, transfer, testimonial, plain-folk, card-stacking, and bandwagon.

In this report, the authors explore methods to detect underlying instances of propaganda devices that might be indicative of broader malign information operations abroad by analyzing hundreds of articles from multiple overseas media outlets. Specifically, the authors explore using large language models (LLMs), which present an opportunity for the detection of overseas propaganda, disinformation, and misinformation targeting foreign audiences. These instances can be indicative of broader attempts by foreign adversaries to undermine U.S. credibility, influence, and alliances abroad.

Unlike traditional methods that rely heavily on specific linguistic indicators or network analysis, LLMs have increasingly large context windows, which offer the ability to incorporate the broader context of the text. The authors assess the performance of these detection tools to gain insights into the capabilities of these technologies to identify overseas propaganda. This represents a first step in preventing malign actors from exploiting the overseas information environment to pursue their strategic goals at the expense of other nations.