# Embedding Poisoning: Bypassing Safety Alignment via Embedding Semantic Shift

arXiv: https://arxiv.org/abs/2509.06338 · HTML: https://mlsanity.com/paper/2509.06338/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Safety alignment and jailbreak attacks, one of 10 directions in LLM alignment / RLHF / preference optimisation.
- Builds on 23 of its 40 references that are in the map.
- First posted on arXiv 8 Sep 2025.

## Placed in

- 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, 23 of 23)

- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [Large Language Models Encode Clinical Knowledge](https://mlsanity.com/paper/2212.13138/) (2022)
- [Jailbreaking Black Box Large Language Models in Twenty Queries](https://mlsanity.com/paper/2310.08419/) (2023)
- [Refusal in Language Models Is Mediated by a Single Direction](https://mlsanity.com/paper/2406.11717/) (2024)
- [Baseline Defenses for Adversarial Attacks Against Aligned Language Models](https://mlsanity.com/paper/2309.00614/) (2023)
- [SmoothLLM: Defending Large Language Models Against Jailbreaking Attacks](https://mlsanity.com/paper/2310.03684/) (2023)
- [GPT-4 Is Too Smart To Be Safe: Stealthy Chat with LLMs via Cipher](https://mlsanity.com/paper/2308.06463/) (2023)
- [Safety-Tuned LLaMAs: Lessons From Improving the Safety of Large Language Models that Follow Instructions](https://mlsanity.com/paper/2309.07875/) (2023)
- [Shadow Alignment: The Ease of Subverting Safely-Aligned Language Models](https://mlsanity.com/paper/2310.02949/) (2023)
- [COLD-Attack: Jailbreaking LLMs with Stealthiness and Controllability](https://mlsanity.com/paper/2402.08679/) (2024)
- [Efficient Adversarial Training in LLMs with Continuous Attacks](https://mlsanity.com/paper/2405.15589/) (2024)
- [Making Them Ask and Answer: Jailbreaking Large Language Models in Few Queries via Disguise and Reconstruction](https://mlsanity.com/paper/2402.18104/) (2024)
- [Soft Prompt Threats: Attacking Safety Alignment and Unlearning in Open-Source LLMs through the Embedding Space](https://mlsanity.com/paper/2402.09063/) (2024)
- [Exploiting LLM Quantization](https://mlsanity.com/paper/2405.18137/) (2024)
- [Adversarial Attacks and Defenses in Large Language Models: Old and New Threats](https://mlsanity.com/paper/2310.19737/) (2023)
- [ASETF: A Novel Method for Jailbreak Attack on LLMs through Translate Suffix Embeddings](https://mlsanity.com/paper/2402.16006/) (2024)
- [Virus: Harmful Fine-tuning Attack for Large Language Models Bypassing Guardrail Moderation](https://mlsanity.com/paper/2501.17433/) (2025)
- [Uncovering Logit Suppression Vulnerabilities in LLM Safety Alignment](https://mlsanity.com/paper/2405.13068/) (2024)
- [Improved Large Language Model Jailbreak Detection via Pretrained Embeddings](https://mlsanity.com/paper/2412.01547/) (2024)
- [Smoothed Embeddings for Robust Language Models](https://mlsanity.com/paper/2501.16497/) (2025)
- [Circumventing Safety Alignment in Large Language Models Through Embedding Space Toxicity Attenuation](https://mlsanity.com/paper/2507.08020/) (2025)

## Abstract (arXiv)

The widespread distribution of Large Language Models (LLMs) through public platforms like Hugging Face introduces significant security challenges. While these platforms perform basic security scans, they often fail to detect subtle manipulations within the embedding layer. This work identifies a novel class of deployment phase attacks that exploit this vulnerability by injecting imperceptible perturbations directly into the embedding layer outputs without modifying model weights or input text. These perturbations, though statistically benign, systematically bypass safety alignment mechanisms and induce harmful behaviors during inference. We propose Search based Embedding Poisoning(SEP), a practical, model agnostic framework that introduces carefully optimized perturbations into embeddings associated with high risk tokens. SEP leverages a predictable linear transition in model responses, from refusal to harmful output to semantic deviation to identify a narrow perturbation window that evades alignment safeguards. Evaluated across six aligned LLMs, SEP achieves an average attack success rate of 96.43% while preserving benign task performance and evading conventional detection mechanisms. Our findings reveal a critical oversight in deployment security and emphasize the urgent need for embedding level integrity checks in future LLM defense strategies.