The report examines how artificial intelligence could reshape clandestine human intelligence operations over the next 1-3 years and what the U.S. Intelligence Community should do to integrate it safely. Its scope is limited to HUMINT espionage, defined as clandestine collection from human sources, and excludes AI applications in cyberoperations, signals intelligence, imagery collection, and all-source analysis. The project draws on workshops beginning in early 2025 with Intelligence Community experts, former HUMINT practitioners, legal experts with IC experience, AI industry representatives, independent research, and consultations based on open, unclassified sources. The report argues that ubiquitous technical surveillance and AI-enabled analysis are making traditional tradecraft increasingly risky, while human sources remain essential for understanding intent and reaching air-gapped or otherwise inaccessible targets. It describes how generative, multimodal, affective, virtual, and agentic AI could support targeting, assessment, development, recruitment, handling, and collection at scale. It also identifies legal ambiguity, data provenance concerns, accountability gaps, and ethical red lines. The policy relevance centers on building human-machine teams governed by Meaningful Human Control, formal guidance, updated Executive Order 12333 procedures, stronger oversight, digital counterintelligence, training, and controlled technical implementation.
