Skip to main content

COMPEL Glossary / excessive-agency

Excessive agency

A failure mode in which an LLM has been wired into tools and permissions whose blast radius exceeds what its supervision and validation logic can safely bound.

What this means in practice

Manifests when prompt-driven action is allowed to cause real-world side effects without human or deterministic confirmation.

Synonyms

excessive permissions , over-privileged agent , LLM06

See also

  • Guardrail — A control placed between the user or environment and an LLM that blocks, rewrites, or classifies content at one of four architectural layers: input filter, policy filter, output filter, or tool-call validator.
  • LLM risk surface — The union of six interacting layers — input, model, output, retrieval, tool, and data — where governance controls must be applied on any LLM-based feature.
  • Human oversight (Art. 14) — Under Regulation (EU) 2024/1689, the provider-designed measures that allow natural persons to understand the capacities and limitations of a high-risk AI system, monitor its operation, and intervene or interrupt it.

Related articles in the Body of Knowledge