{
  "id": 3961,
  "url": "https://arxiv.org/abs/2605.21609v1",
  "title": "CR4T: Rewrite-Based Guardrails for Adolescent LLM Safety",
  "summary": "Large language models (LLMs) are increasingly embedded in adolescent digital environments, mediating information seeking, advice, and emotionally sensitive interactions. Yet existing safety mechanisms remain largely grounded in adult-centric norms and operationalize safety through refusal-oriented suppression. While such approaches may reduce immediate policy violations, they can also create conversational dead-ends, limit constructive guidance, and fail to address the developmental vulnerabilit",
  "authors": "Heajun An, Qi Zhang, Vedanth Achanta, Jin-Hee Cho",
  "category": "research",
  "topics": "regulation,environment",
  "orgs": null,
  "regions": null,
  "published_at": "2026-05-20T18:16:18.000Z",
  "fetched_at": "2026-07-14T16:30:36.745Z",
  "source_slug": "arxiv-ethics",
  "source_name": "arXiv",
  "source_homepage": "https://arxiv.org",
  "ethics_ai_record_url": "https://ethics.ai/record/3961",
  "original_url": "https://arxiv.org/abs/2605.21609v1",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}