The BaseResearcharXiv
Why Do Agent Teams Fail Across Boundaries? A Failure Taxonomy for Agent-to-Agent Systems with Private States
A preprint analyses more than 2,500 recorded runs of agent-to-agent systems across 488 scenarios and derives a taxonomy of 20 failure modes in four categories, from delegation to conduct and fidelity.
Why it matters
The same thing, for different readers.
What it means for you
As personal AI agents negotiate with company agents, failures in discovery, record-keeping and verification could affect real deals.
For lawmakers
Agents acting for different parties raise questions of accountability that single-orchestrator studies do not cover.
For researchers
Supplies 20 failure modes in four categories grounded in recorded traces.
How the finding was reached
In our words, from the source. Weigh the finding by how it was made.
- What was done
- More than 2,500 recorded runs of agent-to-agent systems from different harnesses and models, across 488 scenarios in eight families.
- What it found
- A taxonomy, A^2-Failures, of 20 failure modes in four categories: discovery and delegation, grounding and record consistency, verification, conduct and fidelity.
- What it does not show
- Abstract only; the three cross-cutting observations are truncated in the excerpt.
- Status
- A preprint: not yet peer reviewed
The record
- Questions
- Can AI systems act beyond what people intended or allowed? · How is AI changing childhood, health and daily life?
- People
- Sarath Shekkizhar (Wrote it) · Prathyusha Jwalapuram (Wrote it) · Zixuan Ke (Wrote it)
- What we read
- We have read only the description the publisher gives. No quotations until someone reads the source itself.
- Found by
- The Radar, run of October 15, 2026, via
arxiv-ai-safety-query(tier 1: the source’s own channel or an official record) - Cite this
- arXiv, “Why Do Agent Teams Fail Across Boundaries? A Failure Taxonomy for Agent-to-Agent Systems with Private States”, October 14, 2026. Entry in the Ctrl AI Base, https://ctrlai.com/base/arxiv-2026-10-14-agent-to-agent-failure-taxonomy
Our notesLight lane: written from title, author and abstract excerpt only; not fetched. Authors listed are those named in triage.