Search papers, labs, and topics across Lattice.
This study investigates the risks of prompt injection attacks in agentic systems that utilize persistent memory, focusing on Anthropic Claude Code and OpenAI Codex across multiple model versions. The research reveals that while direct overwriting of memory files with untrusted content is challenging, pre-existing payloads can effectively compromise future sessions, highlighting a significant vulnerability in memory-based architectures. The findings indicate that the presence of persistent memory alters the threat landscape for prompt injection, necessitating new defensive strategies that balance security with agent adaptability.
Existing payloads in memory can compromise future agent behavior, revealing a critical vulnerability in memory-based AI systems.
A growing class of agentic systems maintain persistent state across sessions through memory files, behavioral preferences, and knowledge bases. While this makes agents more useful and self-improving, it also creates a new attack surface for prompt injections in which malicious instructions can be embedded within persistent files and influence future behavior. In this work, we study prompt injection attacks in memory-based agentic systems using a sandboxed synthetic workspace. We evaluate two agentic systems, Anthropic Claude Code and OpenAI Codex, across four models: Claude Haiku 4.5, Claude Opus 4.7, GPT-5.2, and GPT-5.5. Our results show that although it is difficult to make an agent overwrite its own memory files using untrusted external content, payloads already planted in those files can successfully attack current and future sessions. Attack success and payload persistence vary substantially across systems, models, adversarial goals, and multi-session attack sequences. These findings show that persistent memory changes the threat model for prompt injection and motivate defenses that protect memory updates without removing useful agent adaptation.
UW