Analysis · curated 4 Oct 2026
What Is Coerced Tool Use? Definition & Examples
First reported nhimg.org
Coverage timeline
Single-source analysis — first reported, latest, and curated coincide.
Why it matters
Coerced tool use turns an AI agent's legitimate permissions into an attack path that evades single-step policy checks, a trust-boundary problem defenders of agentic systems must address.
A glossary entry from NHI Mgmt Group defines "coerced tool use," an attack in which a prompt steers an AI agent to chain its legitimate tools in a harmful sequence where each individual call appears permitted but the full chain enables data exfiltration, unauthorized changes, or destructive actions. The reference material covers how the technique works, why it is hard to detect, failure conditions, and mitigation guidance such as scoped authorization and tool logging.