{
  "id": 858,
  "url": "https://arxiv.org/abs/2606.18532v1",
  "title": "AI Sandboxes: A Threat Model, Taxonomy, and Measurement Framework",
  "summary": "AI systems are increasingly evaluated in bounded environments that combine isolation, simulation, instrumentation, supervision, and evidence capture. For physical AI, AIoT, and cyber-physical systems, this shift is not a matter of terminology: the system under test may sense, decide, actuate, communicate, and fail through physical processes, networked devices, and human operators. This article develops an assurance-oriented account of AI sandboxes as controlled environments for testing, evaluati",
  "authors": "Inderjeet Singh, Haitham Mahmoud, Andrés Murillo",
  "category": "research",
  "topics": "military-security,environment",
  "orgs": null,
  "regions": null,
  "published_at": "2026-06-16T22:57:24.000Z",
  "fetched_at": "2026-07-14T14:14:50.325Z",
  "source_slug": "arxiv-ethics",
  "source_name": "arXiv",
  "source_homepage": "https://arxiv.org",
  "ethics_ai_record_url": "https://ethics.ai/record/858",
  "original_url": "https://arxiv.org/abs/2606.18532v1",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}