# Activation Steering Induces Emergent Misalignment: Steering Vector Construction and Comprehensive Evaluation

Preprint · Qi Cao, Jian Lou, Meiting Liu, Wenjie Feng, Dan Li, Anh Tuan Luu

- Archive id: arxiv:2606.08682
- Published: 2026-06-07
- arXiv: https://arxiv.org/abs/2606.08682
- Source URL: https://arxiv.org/abs/2606.08682
- PDF: https://arxiv.org/pdf/2606.08682
- Provenance: arxiv

## Abstract

Activation steering (AS) has emerged as a popular inference-time technique for modulating the behavior of large language models (LLMs). AS modulates model behavior in a flexible and lightweight manner by injecting task-specific steering vectors (SVs) into intermediate activations. Meanwhile, recent work has identified emergent misalignment (EM) as an emerging safety concern, whereby finetuning on unsafe examples from a narrow task unexpectedly induces broadly unsafe behavior in unrelated domains. With the increasing popularity of AS, it is pressing yet under-explored to thoroughly investigate whether AS can induce EM. In this paper, we conduct a comprehensive study of activation-steering-induced emergent misalignment. Specifically, (1) we identify a steering vector construction to confirm that SVs derived from a narrow task can indeed induce misalignment across broad unrelated domains. Qualitative measurement further reveals that AS-induced EM may pose more severe safety risks than finetuning-induced EM, i.e., AS elicits more actionable harmful responses with greater semantic coherence than those generated by finetuned models. (2) We characterize the properties of AS-induced EM by analyzing key steering-specific factors, including steering magnitude, the low-rank structure of the steering subspace, and the number of fine-tuning epochs used during steering-vector construction. (3) We evaluate the robustness and sensitivity of AS-induced EM across diverse model families, model scales, target tasks, and intervention layers. Our findings reveal activation steering as a significant yet underexamined source of emergent misalignment and provide an activation-space perspective on emergent misalignment and its safety risks.

_Abstract reproduced under CC0-1.0 (arXiv descriptive metadata)._

---
Space Kitty Archive · https://space-kitty.com · API: https://space-kitty.com/api/v1 · MCP: https://space-kitty.com/mcp · Premium: $0.20/request via x402 (Base Sepolia (testnet)). Metadata: arXiv (CC0), OpenAlex (CC0), Europe PMC (CC BY/CC0 works), ClinicalTrials.gov.