{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-10-26-007",
    "source_story_id": "tmp-story-barc-benign-risk",
    "edition_id": "mp-2026-10-26-morning-0108",
    "edition_url": "https://themachinepress.com/edition/2026-10-26",
    "position": 7,
    "story_type": "dispatch",
    "section": "benchmarks-evals",
    "editorial_classification": "editorial",
    "headline": "The Request Was Benign. The Calendar Leak Was Not.",
    "slug": "the-request-was-benign-the-calendar-leak-was-not",
    "dek": "BARC-Bench tests whether agents infer unstated boundaries from workspace context; even the strongest configuration violated safety on more than a quarter of tasks.",
    "summary": "BARC-Bench tests whether agents infer unstated boundaries from workspace context; even the strongest configuration violated safety on more than a quarter of tasks.",
    "body_text": "BARC-Bench contains 110 simulated workspace tasks in which an ordinary request leaves room for an unsafe implementation. Scheduling an off-boarding discussion, for example, should not expose a resignation through a shared invitation. Four frontier configurations were each tried five times per task. Depending on the system, only 9.1 to 72.7 percent of tasks were completed safely on all five attempts; the strongest still violated a boundary on 27.3 percent of tasks. The benchmark's records and social norms define those judgments, so the figures are not a direct production incident rate. They do show why task completion alone is an inadequate agent metric when critical constraints remain implicit.",
    "why_it_matters": "BARC-Bench tests whether agents infer unstated boundaries from workspace context; even the strongest configuration violated safety on more than a quarter of tasks.",
    "limitations": [
      "Scheduling an off-boarding discussion, for example, should not expose a resignation through a shared invitation.",
      "Depending on the system, only 9.1 to 72.7 percent of tasks were completed safely on all five attempts; the strongest still violated a boundary on 27.3 percent of tasks.",
      "The benchmark's records and social norms define those judgments, so the figures are not a direct production incident rate."
    ],
    "importance": 9,
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-26-007/the-request-was-benign-the-calendar-leak-was-not",
    "json_url": "https://themachinepress.com/story/mp-2026-10-26-007.json",
    "first_published_at": "2026-10-26T09:00:00.000-04:00",
    "modified_at": "2026-10-26T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-10-26-007-001",
        "text": "BARC-Bench tests whether agents infer unstated boundaries from workspace context; even the strongest configuration violated safety on more than a quarter of tasks.",
        "source_ids": [
          "source-2026-10-26-007"
        ],
        "qualification": "Scheduling an off-boarding discussion, for example, should not expose a resignation through a shared invitation."
      }
    ],
    "source_ids": [
      "source-2026-10-26-007"
    ],
    "tags": [
      "agent safety",
      "benchmarks",
      "workspaces",
      "implicit constraints"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-10-26-007",
      "title": "arXiv preprint 2610.32024, original v1",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2610.32024v1",
      "canonical_url": "https://arxiv.org/abs/2610.32024v1",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-10-22T20:53:24.000-04:00",
      "accessed_at": "2026-10-26T08:26:57.002-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-10-26-007-001"
      ]
    }
  ],
  "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 Request Was Benign. The Calendar Leak Was Not.",
    "publisher": "The Machine Press",
    "published_at": "2026-10-26T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-26-007/the-request-was-benign-the-calendar-leak-was-not"
  }
}