# When Words Are Safe But Actions Kill: Probing Physical Jailbreak Beyond Textual Jailbreak in Hidden-State Risk Space

arXiv: https://arxiv.org/abs/2607.15218 · HTML: https://mlsanity.com/paper/2607.15218/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Jailbreaking, one of 12 directions in LLM safety and security.
- Builds on 49 of its 51 references that are in the map.
- First posted on arXiv 16 Jul 2026.

## Placed in

- LLM safety and security / [Jailbreaking](https://mlsanity.com/safety/jailbreaking/)

## Builds on (most cited first, 30 of 49)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [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)
- [Qwen2.5 Technical Report](https://mlsanity.com/paper/2412.15115/) (2024)
- [Universal and Transferable Adversarial Attacks on Aligned Language Models](https://mlsanity.com/paper/2307.15043/) (2023)
- [RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control](https://mlsanity.com/paper/2307.15818/) (2023)
- [Constitutional AI: Harmlessness from AI Feedback](https://mlsanity.com/paper/2212.08073/) (2022)
- [PIQA: Reasoning about Physical Commonsense in Natural Language](https://mlsanity.com/paper/1911.11641/) (2019)
- [Do As I Can, Not As I Say: Grounding Language in Robotic Affordances](https://mlsanity.com/paper/2204.01691/) (2022)
- [G-Eval: NLG Evaluation using GPT-4 with Better Human Alignment](https://mlsanity.com/paper/2303.16634/) (2023)
- [PaLM-E: An Embodied Multimodal Language Model](https://mlsanity.com/paper/2303.03378/) (2023)
- [Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone](https://mlsanity.com/paper/2404.14219/) (2024)
- [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)
- [Sparse Autoencoders Find Highly Interpretable Features in Language Models](https://mlsanity.com/paper/2309.08600/) (2023)
- [Understanding intermediate layers using linear classifier probes](https://mlsanity.com/paper/1610.01644/) (2016)
- [Representation Engineering: A Top-Down Approach to AI Transparency](https://mlsanity.com/paper/2310.01405/) (2023)
- [Inference-Time Intervention: Eliciting Truthful Answers from a Language Model](https://mlsanity.com/paper/2306.03341/) (2023)
- [RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Models](https://mlsanity.com/paper/2009.11462/) (2020)
- [Llama Guard: LLM-based Input-Output Safeguard for Human-AI Conversations](https://mlsanity.com/paper/2312.06674/) (2023)
- [Language Models as Zero-Shot Planners: Extracting Actionable Knowledge for Embodied Agents](https://mlsanity.com/paper/2201.07207/) (2022)
- [Refusal in Language Models Is Mediated by a Single Direction](https://mlsanity.com/paper/2406.11717/) (2024)
- [Steering Language Models With Activation Engineering](https://mlsanity.com/paper/2308.10248/) (2023)
- [Probing Classifiers: Promises, Shortcomings, and Advances](https://mlsanity.com/paper/2102.12452/) (2021)
- [AI2-THOR: An Interactive 3D Environment for Visual AI](https://mlsanity.com/paper/1712.05474/) (2017)
- [Discovering Latent Knowledge in Language Models Without Supervision](https://mlsanity.com/paper/2212.03827/) (2022)
- [The Linear Representation Hypothesis and the Geometry of Large Language Models](https://mlsanity.com/paper/2311.03658/) (2023)
- [Prometheus: Inducing Fine-grained Evaluation Capability in Language Models](https://mlsanity.com/paper/2310.08491/) (2023)
- [Identifying the Risks of LM Agents with an LM-Emulated Sandbox](https://mlsanity.com/paper/2309.15817/) (2023)
- [WildGuard: Open One-Stop Moderation Tools for Safety Risks, Jailbreaks, and Refusals of LLMs](https://mlsanity.com/paper/2406.18495/) (2024)

## Abstract (arXiv)

Large language models (LLMs) increasingly serve as high-level planners for embodied agents, where linguistically benign instructions can become unsafe once grounded in the physical world. We study whether this physically grounded jailbreak is the same safety problem as ordinary textual jailbreak. Through hidden-state direction analysis and random-split null tests, we show that textual jailbreak (TJ) and physical jailbreak (PJ) form separable signals in LLM representations across Qwen2.5-3B/7B/14B/32B, Phi-3.5 and SmolLM2. Building on this separability, we propose PRISM, a single-layer L2-regularized logistic probe over full hidden states. PRISM achieves 86.2--87.7% accuracy on SafeAgentBench with 11.7--13.7% false-positive rates (FPRs), while same-scale LLM judges over-block safe tasks at 24.7--39.0% FPR. To test whether the result survives lexical-shortcut controls, we introduce an interaction-balanced revision of PhysicalJailbreakBench-2K (PJB-2K): a fixed 2,000-row comparison set sampled by label and physical mechanism from a larger object--site construction. On the underlying 10,000-row pool, word-TFIDF and the embedding layer remain at chance (AUC 0.497 and 0.500). At layer 25, selected by an i.i.d. sweep, cell-grouped cross-validation gives PRISM 0.718 AUC, compared with 0.398 for a physics-free label control under the same protocol. On the identical 2,000 comparison rows, these PRISM predictions obtain 0.671 balanced accuracy, while Qwen2.5 judges from 3B to 72B obtain 0.538--0.577 and exhibit high FPR. These results support hidden-state probing as a representation-level method for physical safety beyond text moderation, without relying on the near-perfect scores of shortcut-prone paired templates.