# Membrane: A Self-Evolving Contrastive Safety Memory for LLM Agent Defense

arXiv: https://arxiv.org/abs/2606.05743 · HTML: https://mlsanity.com/paper/2606.05743/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Placed in Agent threats, guardrails and governance, one of 16 directions in LLM agents and tool use.
- Builds on 27 of its 50 references that are in the map.
- First posted on arXiv 4 Jun 2026.

## Placed in

- LLM agents and tool use / [Agent threats, guardrails and governance](https://mlsanity.com/agents/agent-threats-guardrails-and-governance/)
- LLM alignment / RLHF / preference optimisation / [Safety alignment and jailbreak attacks](https://mlsanity.com/alignment/safety-alignment-and-jailbreak-attacks/)
- Synthetic data and self-improvement for LLMs / [Safety of self-evolving systems](https://mlsanity.com/synth/safety-of-self-evolving-systems/)
- LLM safety and security / [Jailbreaking](https://mlsanity.com/safety/jailbreaking/)

## Builds on (most cited first, 27 of 27)

- [Efficient Memory Management for Large Language Model Serving with PagedAttention](https://mlsanity.com/paper/2309.06180/) (2023)
- [Dense Passage Retrieval for Open-Domain Question Answering](https://mlsanity.com/paper/2004.04906/) (2020)
- [Jailbroken: How Does LLM Safety Training Fail?](https://mlsanity.com/paper/2307.02483/) (2023)
- [HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusal](https://mlsanity.com/paper/2402.04249/) (2024)
- [Jailbreaking Black Box Large Language Models in Twenty Queries](https://mlsanity.com/paper/2310.08419/) (2023)
- [MemGPT: Towards LLMs as Operating Systems](https://mlsanity.com/paper/2310.08560/) (2023)
- [Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To!](https://mlsanity.com/paper/2310.03693/) (2023)
- [A-MEM: Agentic Memory for LLM Agents](https://mlsanity.com/paper/2502.12110/) (2025)
- [Tree of Attacks: Jailbreaking Black-Box LLMs Automatically](https://mlsanity.com/paper/2312.02119/) (2023)
- [Passage Re-ranking with BERT](https://mlsanity.com/paper/1901.04085/) (2019)
- [How Johnny Can Persuade LLMs to Jailbreak Them: Rethinking Persuasion to Challenge AI Safety by Humanizing LLMs](https://mlsanity.com/paper/2401.06373/) (2024)
- [WildGuard: Open One-Stop Moderation Tools for Safety Risks, Jailbreaks, and Refusals of LLMs](https://mlsanity.com/paper/2406.18495/) (2024)
- [Jailbreaking Leading Safety-Aligned LLMs with Simple Adaptive Attacks](https://mlsanity.com/paper/2404.02151/) (2024)
- [SmoothLLM: Defending Large Language Models Against Jailbreaking Attacks](https://mlsanity.com/paper/2310.03684/) (2023)
- [A Wolf in Sheep's Clothing: Generalized Nested Jailbreak Prompts can Fool Large Language Models Easily](https://mlsanity.com/paper/2311.08268/) (2023)
- [AutoDAN-Turbo: A Lifelong Agent for Strategy Self-Exploration to Jailbreak LLMs](https://mlsanity.com/paper/2410.05295/) (2024)
- [ShieldAgent: Shielding Agents via Verifiable Safety Policy Reasoning](https://mlsanity.com/paper/2503.22738/) (2025)
- [Aegis2.0: A Diverse AI Safety Dataset and Risks Taxonomy for Alignment of LLM Guardrails](https://mlsanity.com/paper/2501.09004/) (2025)
- [GuardAgent: Safeguard LLM Agents by a Guard Agent via Knowledge-Enabled Reasoning](https://mlsanity.com/paper/2406.09187/) (2024)
- [AGrail: A Lifelong Agent Guardrail with Effective and Adaptive Safety Detection](https://mlsanity.com/paper/2502.11448/) (2025)
- [TrustAgent: Towards Safe and Trustworthy LLM-based Agents](https://mlsanity.com/paper/2402.01586/) (2024)
- [Towards Understanding Jailbreak Attacks in LLMs: A Representation Space Analysis](https://mlsanity.com/paper/2406.10794/) (2024)
- [SelfDefend: LLMs Can Defend Themselves against Jailbreaking in a Practical Manner](https://mlsanity.com/paper/2406.05498/) (2024)
- [Tricking LLMs into Disobedience: Formalizing, Analyzing, and Detecting Jailbreaks](https://mlsanity.com/paper/2305.14965/) (2023)
- [Adversarial D'ej`a Vu: Jailbreak Dictionary Learning for Stronger Generalization to Unseen Attacks](https://mlsanity.com/paper/2510.21910/) (2025)
- [Athena: Safe Autonomous Agents with Verbal Contrastive Learning](https://mlsanity.com/paper/2408.11021/) (2024)
- [ExpGuard: LLM Content Moderation in Specialized Domains](https://mlsanity.com/paper/2603.02588/) (2026)

## Built on by (most cited first, 3 of 3)

- [Yesterday's Shield, Today's Spear: A Self-Evolving Safety Guardrail in Production](https://mlsanity.com/paper/2608.08471/) (2026)
- [VEPA: Verified Evolving Formal Policy for Lifelong LLM Agent Guardrails](https://mlsanity.com/paper/2610.12813/) (2026)
- [Evolve Together, Fall into Place: How Social Scaffolding Shapes Self-Evolving Multi-Agent Systems](https://mlsanity.com/paper/2610.19799/) (2026)

## Abstract (arXiv)

Despite advances in safety alignment, large language models remain vulnerable to continuously evolving jailbreaks. Existing fine-tuned safety classifiers cannot adapt to these evolving attacks, while adaptive memory-based guardrails tend to over-refuse benign queries that resemble stored attacks. We propose Membrane, a self-evolving guardrail built on Contrastive Safety Memory (CSM): each cell pairs the conditions for blocking a harmful query with those for permitting a superficially similar benign request. Without retraining, Membrane evolves CSM by distilling each harmful interaction and its benign counterpart into a contrastive cell indexed by the underlying attack strategy, so that one cell generalizes across topical variants of the same mechanism. At inference, retrieved cells serve as grounding context for precise safety decisions. Across model-level safety on HarmBench and agent-level safety on AgentHarm, Membrane achieves the highest F1 on all six modern jailbreak attacks. Notably, benign refusal on AgentHarm stays at 7-14%, well below the 28-85% range of prior guards. Memory cells also retain 87-88% F1 under cross-attack transfer and remain stable under memory poisoning.