{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-10-24-001",
    "source_story_id": "tmp-lead-post-audit-mirage",
    "edition_id": "mp-2026-10-24-morning-0106",
    "edition_url": "https://themachinepress.com/edition/2026-10-24",
    "position": 1,
    "story_type": "lead",
    "section": "safety-security",
    "editorial_classification": "editorial",
    "headline": "The Audit Saw the Same Evidence. Deployment Did Not.",
    "slug": "the-audit-saw-the-same-evidence-deployment-did-not",
    "dek": "Matched safe and harmful agent lifecycles can look identical to an offline verifier, forcing monitors to distinguish a valid calculation from an answer the evidence cannot support.",
    "summary": "Matched safe and harmful agent lifecycles can look identical to an offline verifier, forcing monitors to distinguish a valid calculation from an answer the evidence cannot support.",
    "body_text": "An offline verifier can calculate correctly and still be unable to identify what will happen after an agent update reaches deployment. The authors construct matched safe and harmful lifecycles that expose identical evidence at the audit boundary while an isolated scorer records opposite outcomes. Their theorem turns that observational equivalence into an error-versus-abstention trade-off. In executable authorization, scheduling and shared-capacity environments, a monitor that receives fresh group-labelled outcomes can sometimes recover the missing distinction, but remains undecided for rare groups, small effects or missing observations. The result is a benchmark and formal boundary, not proof that every offline audit fails.",
    "why_it_matters": "Matched safe and harmful agent lifecycles can look identical to an offline verifier, forcing monitors to distinguish a valid calculation from an answer the evidence cannot support.",
    "limitations": [
      "The result is a benchmark and formal boundary, not proof that every offline audit fails."
    ],
    "importance": 10,
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-24-001/the-audit-saw-the-same-evidence-deployment-did-not",
    "json_url": "https://themachinepress.com/story/mp-2026-10-24-001.json",
    "first_published_at": "2026-10-24T09:00:00.000-04:00",
    "modified_at": "2026-10-24T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-10-24-001-001",
        "text": "Matched safe and harmful agent lifecycles can look identical to an offline verifier, forcing monitors to distinguish a valid calculation from an answer the evidence cannot support.",
        "source_ids": [
          "source-2026-10-24-001"
        ],
        "qualification": "The result is a benchmark and formal boundary, not proof that every offline audit fails."
      }
    ],
    "source_ids": [
      "source-2026-10-24-001"
    ],
    "tags": [
      "agent safety",
      "offline audits",
      "monitoring",
      "identifiability"
    ],
    "image_url": "https://themachinepress.com/issues/2026-10-24/lead-post-audit-mirage.png",
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-10-24-001",
      "title": "arXiv preprint 2610.31130, original v1",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2610.31130v1",
      "canonical_url": "https://arxiv.org/abs/2610.31130v1",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-10-22T07:50:38.000-04:00",
      "accessed_at": "2026-10-24T08:23:48.747-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-10-24-001-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "The Audit Saw the Same Evidence. Deployment Did Not.",
    "publisher": "The Machine Press",
    "published_at": "2026-10-24T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-24-001/the-audit-saw-the-same-evidence-deployment-did-not"
  }
}