A Comparative Analysis of AI-Generated Intelligence Reports

A closed model writes the better assessment. An auditable one writes the more defensible assessment, and that matters more.

The paper runs the same analytic problem (domestic violent extremist threats to federal buildings during ongoing protests) through two very different tools: GPT-4, a closed-source frontier model, and Recorded Future's AI-enabled threat intelligence platform, which combines open-source models with proprietary aggregation and NLP.

The full document is available on request, so please get in touch.