An AI Agent Execution Environment to Safeguard User Data

arXiv:2604.19657 · cs.CR, cs.AI, cs.OS · Submitted 2026-04-21 · Read on arXiv

cs.CR, cs.AI, cs.OS

Submitted: 2026-04-21

Updated: 2026-09-24

Code: https://github.com/DonutShinobu/claude-code-fork

Project page: https://openai.github.io/openai-agents-python/guardrails

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Terminology

• Benchmark — 17× in this paper · explained in 2 episodes
• AI Agents — 11× in this paper · explained in 2 episodes
• Core — 8× in this paper · explained in 2 episodes
• Information Flow Control — 7× in this paper
• Model Context Protocol — 7× in this paper
• Benchmarking — 5× in this paper · explained in 3 episodes
• Agentic System — 4× in this paper · explained in 2 episodes
• Agentic AI — 3× in this paper · explained in 1 episode
• Accuracy — 2× in this paper · explained in 1 episode
• Adversarial Attacks — 2× in this paper · explained in 1 episode
• Social Security Number — 2× in this paper
• Accountability — 1× in this paper · explained in 2 episodes
• Agent Interactions — 1× in this paper · explained in 1 episode
• Agentic AI systems — 1× in this paper · explained in 1 episode
• Agentic Workflows — 1× in this paper · explained in 1 episode
• Fine-Tuning — 1× in this paper · explained in 5 episodes
• Guaranteed Accounting for Agent Privacy — 1× in this paper
• Hallucination — 1× in this paper · explained in 3 episodes
• LLM-as-a-Judge — 1× in this paper · explained in 3 episodes
• Model Context Protocol (MCP) — 1× in this paper · explained in 8 episodes
• Network and Distributed System Security Symposium — 1× in this paper
• Perplexity — 1× in this paper · explained in 3 episodes
• Trace — 1× in this paper · explained in 2 episodes

Sources

• AgentDojo: A Dynamic Environment to Evaluate Prompt Injection Attacks and Defenses for LLM Agents
• Memory Injection Attacks on LLM Agents via Query-Only Interaction
• Text-Based Personas for Simulating User Privacy Decisions
• Personalizing Agent Privacy Decisions via Logical Entailment
• Not what you've signed up for: Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection
• Can LLMs Make (Personalized) Access Control Decisions?
• Defending Against Indirect Prompt Injection Attacks With Spotlighting
• Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversations
• Preventing Prompt Injection with Type-Directed Privilege Separation
• Baseline Defenses for Adversarial Attacks Against Aligned Language Models
• Detecting Language Model Attacks with Perplexity
• AirGapAgent: Protecting Privacy-Conscious Conversational Agents
• AgentPoison: Red-teaming LLM Agents via Poisoning Memory or Knowledge Bases
• LlamaFirewall: An open source guardrail system for building secure AI agents
• Securing AI Agents with Information-Flow Control
• Defeating Prompt Injections by Design
• WebGPT: Browser-assisted question-answering with human feedback
• The Attacker Moves Second: Stronger Adaptive Attacks Bypass Defenses Against Llm Jailbreaks and Prompt Injections
• Formal Policy Enforcement for Real-World Agentic Systems
• Red Teaming the Mind of the Machine: A Systematic Evaluation of Prompt Injection and Jailbreak Vulnerabilities in LLMs

Related papers

• SoK: AI-Augmented Binary Reversing
• Relaxed Sender Anonymity for CBDC Interbank Settlement: A Zero-Knowledge Approach on Permissioned EVM
• Calibration-Family Overfit: Why Trusted Sabotage Monitors Don't Transfer Across Lineages
• Efficient Fuzzy PSI under One-Sided Assumptions
• Sealing the Audit-Runtime Gap for LLM Skills
• Token Composition: A Graph Based on EVM Logs