AI Paper Atlas

Research map / Multi-agent systems and games / LLM agents

Cross-User Poisoning: User-Task Boundary Failures in Multi-User Collaborative Language Agents

Atharv Singh Patlan, Peiyao Sheng, S Ashwin Hebbar, Prateek Mittal, Pramod Viswanath

NeurIPS 2026 · Multi-agent systems and games › LLM agents · approach: Agent safety and attack benchmarks · 4 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Identifies a user-task boundary problem in multi-user collaborative language agents, where a shared agent must decide which user or task an instruction may govern, creating cross-user poisoning risks.

Problem

Shared language agents observe interleaved messages from multiple users, keep persistent context, and execute tool actions, so they must decide not only whether an instruction is safe but which user or task it is allowed to govern.

Approach

Frames cross-user poisoning as a boundary failure specific to multi-user shared agents, distinct from standard single-user agent safety concerns about whether an instruction is safe.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add task clustering (isolate inferred task context) (ablation)GPT-4.1 shared agent · Slack, Workspace, AirlineASR: 0.0 ASR in all three environmentsImproved
add context augmentation (task-cluster IDs and summaries) (ablation)GPT-4.1 shared agent · Slack, Workspace, AirlineASR: 0.0 in Slack and Workspace, 5.7 in AirlineImproved
add selective summarization of other users' context (ablation)GPT-4.1 shared agent · Slack, Workspace, AirlineASR: 23-48% ASR remains; TSR lowered by 15-25 pointsNo clear effect
add task clustering (ablation)GPT-4.1 shared agent · Slack, Workspace, AirlineTSR: Slack 81.0 -> 64.0; Workspace 72.1 -> 72.8; Airline 46.7 -> 41.1Hurt
add context augmentation (ablation)GPT-4.1 shared agent · Slack, WorkspaceTSR: Slack 81.0 -> 96.2; Workspace 72.1 -> 95.4Improved
add ProtectAI input filter against CUPGPT-4.1 · MURMUR avgASR: 69.4 -> 43.3Mixed
add PromptGuard 2 input filter against CUPGPT-4.1 · MURMURASR: 42.7% ASR retainedMixed
deliver attack objective via another user (CUP) instead of retrieved data (PI)frontier models · Slack and WorkspaceASR: PI ASR below 5%; CUP nontrivial and often highImproved
use instruction-detecting PI defenses on benign messagesDataSentinel, MELON, PromptGuard, GPT-5.4 judge · benign messagesfalse positive rate: 89.06, 100.00, 53.12, 35.53 percent flaggedHurt
use TS-Guard action guard, aggressive threshold (risk >= 0.5)GPT-4.1 · MURMURdetection of CUP sessions: detects 85.4 while disrupting 27.93 of benign sessionsMixed
replace main attacker template with minimal or natural-language template (ablation)CUP on GPT-4.1 · Workspace, Slack, AirlineASR: Slack 81.9 -> 64.4 (natural language)Hurt
increase number of concurrent tasksCUP attacks · MURMUR k up to 5compromised tasks: more tasks compromised as concurrency growsHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy