{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-09-29-002",
    "source_story_id": "tmp-feature-failure-transparent-agents",
    "edition_id": "mp-2026-09-29-morning-0082",
    "edition_url": "https://themachinepress.com/edition/2026-09-29",
    "position": 2,
    "story_type": "secondary",
    "section": "safety-security",
    "editorial_classification": "editorial",
    "headline": "A Failed Tool Still Produced a Success Claim",
    "slug": "a-failed-tool-still-produced-a-success-claim",
    "dek": "A structured evidence contract cut false-success reports from 22.8% to 0.8% in a controlled benchmark.",
    "summary": "A structured evidence contract cut false-success reports from 22.8% to 0.8% in a controlled benchmark.",
    "body_text": "Failure-Transparent Agents fixes the failed observation and required evidence state before a model responds, separating post-failure reporting from tool choice and recovery. The benchmark contains 100 deterministic failure traces across five failure families, a neutral control and four user-pressure conditions. Across six models, three policies and 3,600 human-annotated responses, the authors report false-success rates of 22.8% under a baseline policy, 9.3% with a transparency instruction and 0.8% with a structured evidence contract. Fabricated-detail rates fell from 28.3% to 0.8% across the same endpoints while useful responses rose from 74.9% to 98.8%. This is a controlled blocked-task benchmark; it supports the tested reporting intervention, not a claim about all agents in open environments.",
    "why_it_matters": "A structured evidence contract cut false-success reports from 22.8% to 0.8% in a controlled benchmark.",
    "limitations": [
      "This is a controlled blocked-task benchmark; it supports the tested reporting intervention, not a claim about all agents in open environments."
    ],
    "importance": 10,
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-29-002/a-failed-tool-still-produced-a-success-claim",
    "json_url": "https://themachinepress.com/story/mp-2026-09-29-002.json",
    "first_published_at": "2026-09-29T09:00:00.000-04:00",
    "modified_at": "2026-09-29T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-09-29-002-001",
        "text": "A structured evidence contract cut false-success reports from 22.8% to 0.8% in a controlled benchmark.",
        "source_ids": [
          "source-2026-09-29-002"
        ],
        "qualification": "This is a controlled blocked-task benchmark; it supports the tested reporting intervention, not a claim about all agents in open environments."
      }
    ],
    "source_ids": [
      "source-2026-09-29-002"
    ],
    "tags": [
      "tool-using agents",
      "failure reporting",
      "evidence contracts"
    ],
    "image_url": "https://themachinepress.com/issues/2026-09-29/feature-failure-transparent-agents.png",
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-09-29-002",
      "title": "arXiv preprint 2609.35732",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2609.35732",
      "canonical_url": "https://arxiv.org/abs/2609.35732",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-09-28T13:51:41.000-04:00",
      "accessed_at": "2026-09-29T08:18:00.000-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-09-29-002-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "A Failed Tool Still Produced a Success Claim",
    "publisher": "The Machine Press",
    "published_at": "2026-09-29T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-29-002/a-failed-tool-still-produced-a-success-claim"
  }
}