# LiSA: Lifelong Safety Adaptation via Conservative Policy Induction

arXiv: https://arxiv.org/abs/2605.14454 · HTML: https://mlsanity.com/paper/2605.14454/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Placed in Prompt injection and agent security, one of 12 directions in LLM safety and security.
- Builds on 37 of its 46 references that are in the map.
- First posted on arXiv 14 May 2026.

## Placed in

- LLM safety and security / [Prompt injection and agent security](https://mlsanity.com/safety/prompt-injection-and-agent-security/)

## Builds on (most cited first, 30 of 37)

- [Reflexion: Language Agents with Verbal Reinforcement Learning](https://mlsanity.com/paper/2303.11366/) (2023)
- [τ-bench: A Benchmark for Tool-Agent-User Interaction in Real-World Domains](https://mlsanity.com/paper/2406.12045/) (2024)
- [Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversations](https://mlsanity.com/paper/2312.06674/) (2023)
- [Refusal in Language Models Is Mediated by a Single Direction](https://mlsanity.com/paper/2406.11717/) (2024)
- [WildGuard: Open One-Stop Moderation Tools for Safety Risks, Jailbreaks, and Refusals of LLMs](https://mlsanity.com/paper/2406.18495/) (2024)
- [AgentHarm: A Benchmark for Measuring Harmfulness of LLM Agents](https://mlsanity.com/paper/2410.09024/) (2024)
- [WorkArena: How Capable Are Web Agents at Solving Common Knowledge Work Tasks?](https://mlsanity.com/paper/2403.07718/) (2024)
- [Memory in the Age of AI Agents](https://mlsanity.com/paper/2512.13564/) (2025)
- [ReasoningBank: Scaling Agent Self-Evolving with Reasoning Memory](https://mlsanity.com/paper/2509.25140/) (2025)
- [Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety](https://mlsanity.com/paper/2507.11473/) (2025)
- [ShieldGemma: Generative AI Content Moderation Based on Gemma](https://mlsanity.com/paper/2407.21772/) (2024)
- [Can LLMs Keep a Secret? Testing Privacy Implications of Language Models via Contextual Integrity Theory](https://mlsanity.com/paper/2310.17884/) (2023)
- [Synapse: Trajectory-as-Exemplar Prompting with Memory for Computer Control](https://mlsanity.com/paper/2306.07863/) (2023)
- [PrivacyLens: Evaluating Privacy Norm Awareness of Language Models in Action](https://mlsanity.com/paper/2409.00138/) (2024)
- [GuardAgent: Safeguard LLM Agents by a Guard Agent via Knowledge-Enabled Reasoning](https://mlsanity.com/paper/2406.09187/) (2024)
- [In Prospect and Retrospect: Reflective Memory Management for Long-term Personalized Dialogue Agents](https://mlsanity.com/paper/2503.08026/) (2025)
- [GuardReasoner: Towards Reasoning-based LLM Safeguards](https://mlsanity.com/paper/2501.18492/) (2025)
- [Gemini Embedding: Generalizable Embeddings from Gemini](https://mlsanity.com/paper/2503.07891/) (2025)
- [AGrail: A Lifelong Agent Guardrail with Effective and Adaptive Safety Detection](https://mlsanity.com/paper/2502.11448/) (2025)
- [Refuse Whenever You Feel Unsafe: Improving Safety in LLMs via Decoupled Refusal Training](https://mlsanity.com/paper/2407.09121/) (2024)
- [PolyGuard: A Multilingual Safety Moderation Tool for 17 Languages](https://mlsanity.com/paper/2504.04377/) (2025)
- [Continual Learning on Noisy Data Streams via Self-Purified Replay](https://mlsanity.com/paper/2110.07735/) (2021)
- [Personalized Safety in LLMs: A Benchmark and A Planning-Based Agent Approach](https://mlsanity.com/paper/2505.18882/) (2025)
- [ConVerse: Benchmarking Contextual Safety in Agent-to-Agent Conversations](https://mlsanity.com/paper/2511.05359/) (2025)
- [LifeTox: Unveiling Implicit Toxicity in Life Advice](https://mlsanity.com/paper/2311.09585/) (2023)
- [Handling Out-of-Distribution Data: A Survey](https://mlsanity.com/paper/2507.21160/) (2025)
- [CausalArmor: Efficient Indirect Prompt Injection Guardrails via Causal Attribution](https://mlsanity.com/paper/2602.07918/) (2026)
- [PSG-Agent: Personality-Aware Safety Guardrail for LLM-based Agents](https://mlsanity.com/paper/2509.23614/) (2025)
- [Privacy Reasoning in Ambiguous Contexts](https://mlsanity.com/paper/2506.12241/) (2025)
- [AgentGuardian: Learning Access Control Policies to Govern AI Agent Behavior](https://mlsanity.com/paper/2601.10440/) (2026)

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

- [When the Agent Becomes the Kernel: A Systematization of Security on the Path to AI-Native Operating Systems](https://mlsanity.com/paper/2609.23700/) (2026)
- [VEPA: Verified Evolving Formal Policy for Lifelong LLM Agent Guardrails](https://mlsanity.com/paper/2610.12813/) (2026)
- [AuthGateBench: Benchmarking LLMs as Authorization Gate](https://mlsanity.com/paper/2610.16426/) (2026)

## Abstract (arXiv)

As AI agents move from chat interfaces to systems that read private data, call tools, and execute multi-step workflows, guardrails become a last line of defense against concrete deployment harms. In these settings, guardrail failures are no longer merely answer-quality errors: they can leak secrets, authorize unsafe actions, or block legitimate work. The hardest failures are often contextual: whether an action is acceptable depends on local privacy norms, organizational policies, and user expectations that resist pre-deployment specification. This creates a practical gap: guardrails must adapt to their own operating environments, yet deployment feedback is typically limited to sparse, noisy user-reported failures, and repeated fine-tuning is often impractical. To address this gap, we propose LiSA (Lifelong Safety Adaptation), a conservative policy induction framework that improves a fixed base guardrail through structured memory. LiSA converts occasional failures into reusable policy abstractions so that sparse reports can generalize beyond individual cases, adds conflict-aware local rules to prevent overgeneralization in mixed-label contexts, and applies evidence-aware confidence gating via a posterior lower bound, so that memory reuse scales with accumulated evidence rather than empirical accuracy alone. Across PrivacyLens+, ConFaide+, and AgentHarm, LiSA consistently outperforms strong memory-based baselines under sparse feedback, remains robust under noisy user feedback even at 20% label-flip rates, and pushes the latency--performance frontier beyond backbone model scaling. Ultimately, LiSA offers a practical path to secure AI agents against the unpredictable long tail of real-world edge risks.