Author

Satoshi Matsuoka

h-index 1 View on Semantic Scholar

We have 1 of 5 papers

We haven’t gathered this author’s papers yet. Follow them and we’ll fetch their work.

Not the right person? Other researchers publish under this name.

Review Aug 2026

Compositional Threat Analysis of Latent Compromise in LLM Agent Systems: The Order 66 Scenario

This paper introduces a compositional model explaining why no component is catastrophic alone, yet their conjunction can produce correlated destructive action, and separates three population-reach routes from a common core of dormancy, activation, authority, reachable targets, and failed recovery.

Satoshi Matsuoka · 0 citations