A benchmark of 85 manually curated workplace scenarios reveals that multi-user AI agent systems suffer high rates of contextual integrity violations across outputs, inter-agent communication, and shared memory.
Airgapagent: Protecting privacy-conscious conversational agents
5 Pith papers cite this work, alongside 10 external citations. Polarity classification is still indexing.
5
Pith papers citing it
10
external citations · external index
citation-role summary
method 1
citation-polarity summary
roles
method 1polarities
use method 1representative citing papers
Free-text pseudonymization provides a stronger privacy-utility tradeoff than suppression or generalization for LLM-drafted messages, and single-message evaluations underestimate leakage that appears under follow-up questions.
LLMs exhibit 20-40% lower recall on ambiguous human names for PII detection, worsening under prompt injections, as shown via the new AmBench benchmark.
SELFCI uses complementary self-distillation with two reverse KL divergences to align LLMs to contextual integrity while preserving utility, outperforming RL baselines like GRPO in agentic settings.
A synthesis of 247 papers on LLM agent security identifies prompt injection and tool hijacking as dominant threats, notes weakly compositional defenses, and argues for trust boundaries and realistic evaluations.
citing papers explorer
-
PiSAs: Benchmarking Contextual Integrity in Multi-User Agentic Systems
A benchmark of 85 manually curated workplace scenarios reveals that multi-user AI agent systems suffer high rates of contextual integrity violations across outputs, inter-agent communication, and shared memory.
-
Say Something Else: Rethinking Contextual Privacy as Information Sufficiency
Free-text pseudonymization provides a stronger privacy-utility tradeoff than suppression or generalization for LLM-drafted messages, and single-message evaluations underestimate leakage that appears under follow-up questions.
-
Can Large Language Models Really Recognize Your Name?
LLMs exhibit 20-40% lower recall on ambiguous human names for PII detection, worsening under prompt injections, as shown via the new AmBench benchmark.
-
It Takes Two: Complementary Self-Distillation for Contextual Integrity in LLMs
SELFCI uses complementary self-distillation with two reverse KL divergences to align LLMs to contextual integrity while preserving utility, outperforming RL baselines like GRPO in agentic settings.
-
Toward Secure LLM Agents: Threat Surfaces, Attacks, Defenses, and Evaluation
A synthesis of 247 papers on LLM agent security identifies prompt injection and tool hijacking as dominant threats, notes weakly compositional defenses, and argues for trust boundaries and realistic evaluations.