Frontier AI Models for Analysts

Can political and security affairs analysts use artificial intelligence to improve their tradecraft?

Many governments are increasingly using (often-customised) large language models for intelligence synthesis and analysis. These models have also found use in the broader policy world, often as war-gaming and strategic warning tools.

Building on these trends, Tarqeq recently conducted two experiments using frontier models. The first explored the extent to which they can generate granular scenarios, develop and assess futures against fixed analytical frameworks, and — with guidance from formal methods such as causal graphs — remove inconsistencies in narratives. The second experiment involved a model carefully examining an intelligence memorandum in order to parse its contents, inferring the broad arguments presented in the memo, and creating a well-defined formal representation of its findings. The results from both are encouraging.

At the same time, there are significant obstacles when it comes to widespread adoption of LLMs for research and analysis.

Read the full article in The Interpreter