{
  "id": 4016,
  "url": "https://arxiv.org/abs/2605.19940v1",
  "title": "Robotics-Inspired Guardrails for Foundation Models in Socially Sensitive Domains",
  "summary": "Foundation models are increasingly deployed in socially sensitive domains such as education, mental health, and caregiving, where failures are often cumulative and context-dependent. Existing guardrail approaches -- ranging from training-time alignment to prompting, decoding constraints, and post-hoc moderation -- primarily provide empirical risk reduction rather than enforceable behavioral guarantees, and largely treat safety as a property of individual outputs rather than interaction trajector",
  "authors": "Rebecca Ramnauth, Drazen Brscic, Brian Scassellati",
  "category": "research",
  "topics": "safety-alignment,healthcare,children-education,agents-autonomy",
  "orgs": null,
  "regions": null,
  "published_at": "2026-05-19T15:00:06.000Z",
  "fetched_at": "2026-07-14T16:30:41.582Z",
  "source_slug": "arxiv-ethics",
  "source_name": "arXiv",
  "source_homepage": "https://arxiv.org",
  "ethics_ai_record_url": "https://ethics.ai/record/4016",
  "original_url": "https://arxiv.org/abs/2605.19940v1",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}