Where Rules End and Judges Begin: Measuring the Judgment Boundary in Multi-Agent Systems Security
Factual evidence
What the source reports
Researchers introduced DEFER1, a multi-agent security framework combining deterministic rule cascades with LLM judgment for agent defense.
Inspect the evidence
- Inclusion basis
- Enterprise AI
- Publisher and source type
- arXiv cs.CL — Computation and Language · RESEARCH
- Published by source
- 7 October 2026
- Collected by OneBench
- 8 Oct 2026, 03:01 UK
Stored source excerpt
arXiv:2610.07657v1 Announce Type: cross Abstract: LLM-based multi-agent systems (MAS) engage tools, share memory, and delegate tasks, often encountering adversarial content. Current defenses for MAS are…
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
Hybrid rule-and-judgment defense architectures reduce audit overhead and operational risks in enterprise multi-agent deployments.
Do what
Review agentic security framework architecture against the proposed deterministic-first defense concepts during agent workflow design.