Metraipolis › archive › dc_0502b8cf323c

The Guard That Cried Wolf: How Scary Words Make Agent Guardrails Refuse Legitimate Actions — arXiv:2608.27009

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.27009
License (as archived)
CC0-1.0
Archived
2026-10-17T15:55:00.190Z
SHA-256 of the archived text
23af6cc63633087a735221ced8d0c09ec0d39631bdc054dcb5067487517167a0
Stable URL of this page
https://metraipolis.com/documento/dc_0502b8cf323c

How to cite

The Guard That Cried Wolf: How Scary Words Make Agent Guardrails Refuse Legitimate Actions — arXiv:2608.27009 — arXiv (en), https://arxiv.org/abs/2608.27009 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_0502b8cf323c (Metraipolis archive dc_0502b8cf323c, sha256 23af6cc63633087a735221ced8d0c09ec0d39631bdc054dcb5067487517167a0)
BibTeX
@misc{dc0502b8cf323c,
  title = {The Guard That Cried Wolf: How Scary Words Make Agent Guardrails Refuse Legitimate Actions — arXiv:2608.27009},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.27009},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_0502b8cf323c; sha256 23af6cc63633087a735221ced8d0c09ec0d39631bdc054dcb5067487517167a0},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.27009. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Agent guardrails are checks that approve or refuse each action before an LLM executes it. Sometimes they refuse requests that are genuinely safe. This over-safety blocks deployment when a guardrail refuses an authorized task. Evaluating over-safety is hard: at the boundary an authorized action resembles an unauthorized one, and the safe-versus-unsafe label is a choice of authorization policy, not fixed by the action alone. We argue it therefore requires a benchmark that does not yet exist, one that maps the decision boundary of an ideal guardrail. Harvesting such a benchmark from real data is impractical: boundary cases are hard to collect, their labels hard to verify. The gap is real, so we construct Cautious Bench, the first benchmark to make over-safety the construct for agent guardrails; it codesigns each sample and its label with a stated authorization policy. A build-time gate re-derives every example to certify it, so each label is a mechanical consequence of the policy rather than an annotator's per-sample verdict, a reference against which researchers can measure real guardrails. The benchmark renders 756 Decidable benign/twin pairs, each under three object-name types (2,268 measured pairs), and 40 Undecidable pairs reported separately. Measuring six guardrails from five designs, we find a name-superstition effect: each over-refuses an authorized action more often under a scary-looking object name than a benign one. Since only the object name varies in the aforementioned contrast experiments, the deviation is the name's doing: the guardrails read the surface label, not the authorization context.

Tema degli esploratori: fiducia digitale e crittografia (retro-catalogo arXiv: annunci del 2026-08-28 in cs.CR).
Categorie arXiv: Cryptography and Security (cs.CR).
Pagina del lavoro: https://arxiv.org/abs/2608.27009 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-17T15:31:38.272Z dalla pagina https://arxiv.org/catchup/cs.CR/2026-08-28?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).