# MCP Safety Training: Learning to Refuse Falsely Benign MCP Exploits using Improved Preference Alignment

arXiv: https://arxiv.org/abs/2505.23634 · HTML: https://mlsanity.com/paper/2505.23634/

## 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 26 of its 45 references that are in the map.
- First posted on arXiv 29 May 2025.

## 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/)
- LLM safety and security / [Safety alignment, refusal and guardrails](https://mlsanity.com/safety/safety-alignment-refusal-and-guardrails/)

## Builds on (most cited first, 26 of 26)

- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks](https://mlsanity.com/paper/2005.11401/) (2020)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning](https://mlsanity.com/paper/2501.12948/) (2025)
- [Qwen2.5 Technical Report](https://mlsanity.com/paper/2412.15115/) (2024)
- [QLoRA: Efficient Finetuning of Quantized LLMs](https://mlsanity.com/paper/2305.14314/) (2023)
- [The Curious Case of Neural Text Degeneration](https://mlsanity.com/paper/1904.09751/) (2019)
- [Refusal in Language Models Is Mediated by a Single Direction](https://mlsanity.com/paper/2406.11717/) (2024)
- [JailbreakBench: An Open Robustness Benchmark for Jailbreaking Large Language Models](https://mlsanity.com/paper/2404.01318/) (2024)
- [ToxiGen: A Large-Scale Machine-Generated Dataset for Adversarial and Implicit Hate Speech Detection](https://mlsanity.com/paper/2203.09509/) (2022)
- [Statistical Rejection Sampling Improves Preference Optimization](https://mlsanity.com/paper/2309.06657/) (2023)
- [Self-Play Preference Optimization for Language Model Alignment](https://mlsanity.com/paper/2405.00675/) (2024)
- [Purple Llama CyberSecEval: A Secure Coding Benchmark for Language Models](https://mlsanity.com/paper/2312.04724/) (2023)
- [Provably Robust DPO: Aligning Language Models with Noisy Feedback](https://mlsanity.com/paper/2403.00409/) (2024)
- [LlamaFirewall: An open source guardrail system for building secure AI agents](https://mlsanity.com/paper/2505.03574/) (2025)
- [RedCode: Risky Code Execution and Generation Benchmark for Code Agents](https://mlsanity.com/paper/2411.07781/) (2024)
- [Noise Contrastive Alignment of Language Models with Explicit Rewards](https://mlsanity.com/paper/2402.05369/) (2024)
- [MCP Safety Audit: LLMs with the Model Context Protocol Allow Major Security Exploits](https://mlsanity.com/paper/2504.03767/) (2025)
- [Surgical, Cheap, and Flexible: Mitigating False Refusal in Language Models via Single Vector Ablation](https://mlsanity.com/paper/2410.03415/) (2024)
- [Anchored Preference Optimization and Contrastive Revisions: Addressing Underspecification in Alignment](https://mlsanity.com/paper/2408.06266/) (2024)
- [Binary Classifier Optimization for Large Language Model Alignment](https://mlsanity.com/paper/2404.04656/) (2024)
- [Towards Efficient Exact Optimization of Language Model Alignment](https://mlsanity.com/paper/2402.00856/) (2024)
- [MCP Guardian: A Security-First Layer for Safeguarding MCP-Based AI System](https://mlsanity.com/paper/2504.12757/) (2025)
- [Distributional Preference Alignment of LLMs via Optimal Transport](https://mlsanity.com/paper/2406.05882/) (2024)
- [RAG LLMs are Not Safer: A Safety Analysis of Retrieval-Augmented Generation for Large Language Models](https://mlsanity.com/paper/2504.18041/) (2025)

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

- [A Survey of Context Engineering for Large Language Models](https://mlsanity.com/paper/2507.13334/) (2025)
- [MCP Security Bench (MSB): Benchmarking Attacks Against Model Context Protocol in LLM Agents](https://mlsanity.com/paper/2510.15994/) (2025)
- [ShareLock: A Stealthy Multi-Tool Threshold Poisoning Attack Against MCP](https://mlsanity.com/paper/2606.27027/) (2026)

## Abstract (arXiv)

The model context protocol (MCP) has been widely adapted as an open standard enabling the seamless integration of generative AI agents. However, recent work has shown the MCP is susceptible to retrieval-based "falsely benign" attacks (FBAs), allowing malicious system access and credential theft, but requiring that users download compromised files directly to their systems. Herein, we show that the threat model of MCP-based attacks is significantly broader than previously thought, i.e., attackers need only post malicious content online to deceive MCP agents into carrying out their attacks on unsuspecting victims' systems. To improve alignment guardrails against such attacks, we introduce a new MCP dataset of FBAs and (truly) benign samples to explore the effectiveness of direct preference optimization (DPO) for the refusal training of large language models (LLMs). While DPO improves model guardrails against such attacks, we show that the efficacy of refusal learning varies drastically depending on the model's original post-training alignment scheme--e.g., GRPO-based LLMs learn to refuse extremely poorly. Thus, to further improve FBA refusals, we introduce Retrieval Augmented Generation for Preference alignment (RAG-Pref), a novel preference alignment strategy based on RAG. We show that RAG-Pref significantly improves the ability of LLMs to refuse FBAs, particularly when combined with DPO alignment, thus drastically improving guardrails against MCP-based attacks.