Agentic Cyber Explorerby Fide AI · PreviewPreview
Chronicle/Defense & research

PNNL uses a Claude-based agent to speed adversary emulation against a water treatment plant model

DefensePaperSignificance assistant-drafted

Anthropic reports that Pacific Northwest National Laboratory built a scaffold around Claude Sonnet 4 to automate adversary emulation against a high-fidelity cyber-physical model of a water treatment plant used for CISA. PNNL estimates attack reconstruction took three hours instead of multiple weeks; in one run the model switched to a different known privilege-escalation technique when a provided tool failed.

Why it matters

Faster adversary emulation lets critical-infrastructure defenders re-test controls more often, while the model's improvisation shows why such agents need tight scoping.

Key facts

As stated in the sources, with where to find them.

  • PNNL estimates the Claude scaffold allowed attack reconstruction in three hours instead of multiple weeks.Section 'Using AI to speed up adversary emulation'
  • The simulation, run in summer 2025, used Claude Sonnet 4 against a Control Environment Laboratory Resource water treatment platform operated for CISA.Same section
  • When a scaffold tool for bypassing Windows UAC failed, Claude used a different known bypass technique.Same section

Findings that cite this record

No tracked finding cites this record yet.

Sources

Related records