Bad Memory: Evaluating Prompt Injection Risks from Memory in Agentic Systems
For developers and users of agentic systems with persistent memory, this work highlights a new attack surface that changes the threat model for prompt injection, motivating defenses that secure memory updates without hindering adaptation.
The paper investigates prompt injection risks in memory-based agentic systems, finding that while agents resist overwriting their own memory, pre-planted malicious payloads in memory files can influence behavior across sessions. Attack success varies by system, model, and adversarial goal.
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.