{
  "id": 4869,
  "url": "https://arxiv.org/abs/2605.05630v2",
  "title": "One Turn Too Late: Response-Aware Defense Against Hidden Malicious Intent in Multi-Turn Dialogue",
  "summary": "Hidden malicious intent in multi-turn dialogue poses a growing threat to deployed large language models (LLMs). Rather than exposing a harmful objective in a single prompt, increasingly capable attackers can distribute their intent across multiple benign-looking turns. Recent studies show that even modern commercial models with advanced guardrails remain vulnerable to such attacks despite advances in safety alignment and external guardrails. In this work, we address this challenge by detecting t",
  "authors": "Xinjie Shen, Rongzhe Wei, Peizhi Niu, Haoyu Wang, Ruihan Wu, Eli Chien et al.",
  "category": "research",
  "topics": "safety-alignment,military-security",
  "orgs": null,
  "regions": null,
  "published_at": "2026-05-07T03:35:31.000Z",
  "fetched_at": "2026-07-14T16:31:21.930Z",
  "source_slug": "arxiv-ethics",
  "source_name": "arXiv",
  "source_homepage": "https://arxiv.org",
  "ethics_ai_record_url": "https://ethics.ai/record/4869",
  "original_url": "https://arxiv.org/abs/2605.05630v2",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}