DEPUTYBENCH: Evaluating Contextual Safety in Enterprise Agents
Factual evidence
What the source reports
Researchers introduced DEPUTYBENCH to evaluate enterprise AI agents' context-dependent permission compliance across 12 models.
Inspect the evidence
- Inclusion basis
- Enterprise AI
- Publisher and source type
- arXiv cs.LG — Machine Learning · RESEARCH
- Published by source
- 11 November 2026
- Collected by OneBench
- 12 Nov 2026, 02:02 UK
- Original headline
- DEPUTYBENCH: Evaluating Contextual Safety in Enterprise Agents ↗
Stored source excerpt
arXiv:2611.13355v1 Announce Type: cross Abstract: Enterprise agents must determine whether ordinary requests are permitted by organizational records and policies while continuing legitimate work. We introduce…
Short excerpt from the collected text, not the full source. Use the source link to read it in context.
The factual summary is a OneBench synthesis, not a quotation or independent verification. Collection time is not publication time. Open the source for its full context; related reporting can share the same underlying announcement.
OneBench interpretation
Institutional assessment
So what
Enterprise agents frequently fail policy enforcement when authorization depends on dynamic organizational context rather than explicit system prompts.
Do what
Review agent evaluation framework design with the team responsible for model risk to assess contextual policy boundaries.