{
  "id": 302,
  "url": "https://arxiv.org/abs/2607.02714v2",
  "title": "Not All Refusals Are Equal: How Safety Alignment Fails Cybersecurity at Scale",
  "summary": "There is no doubt that safety alignment is an essential step in LLM training. However, conceptually it does not distinguish between various domains and the level of potential harm of a query, which creates significant complications in the fields like cyber security, where a model should not be constrained by its safety circuits to accomplish the goals of legitimate, authorized operations. In this work, we share our findings from a large scale abliteration experiment on 24 open-source LLMs and sh",
  "authors": "Vadym Hadetskyi, Dario Pasquini, Artem Sorokin",
  "category": "research",
  "topics": "safety-alignment,military-security",
  "orgs": null,
  "regions": null,
  "published_at": "2026-07-02T19:05:07.000Z",
  "fetched_at": "2026-07-14T14:14:28.433Z",
  "source_slug": "arxiv-ethics",
  "source_name": "arXiv",
  "source_homepage": "https://arxiv.org",
  "ethics_ai_record_url": "https://ethics.ai/record/302",
  "original_url": "https://arxiv.org/abs/2607.02714v2",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}