{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-09-24-003",
    "source_story_id": "tmp-story-eu-ai-act-systemic-risk-dashboard",
    "edition_id": "mp-2026-09-24-morning-0077",
    "edition_url": "https://themachinepress.com/edition/2026-09-24",
    "position": 3,
    "story_type": "dispatch",
    "section": "policy-regulation",
    "editorial_classification": "editorial",
    "headline": "Worst-Case Scoring Moved AI Risk Ratings by Up to 37 Points",
    "slug": "worst-case-scoring-moved-ai-risk-ratings-by-up-to-37-points",
    "dek": "An open dashboard lets users trace four EU Code of Practice risk categories back to 19 benchmarks.",
    "summary": "An open dashboard lets users trace four EU Code of Practice risk categories back to 19 benchmarks.",
    "body_text": "The Systemic Risk Index organizes 19 public benchmarks into CBRN, cyber offense, harmful manipulation and loss-of-control categories from the EU GPAI Code of Practice. Across 18 models, changing from average to worst-case aggregation lowered scores by 14 to 37 points, showing how a summary choice can hide weak areas. A blind audit found 83% of sampled benchmark transformations preserved the original harm. The tool is an open evidence interface, not an official EU compliance determination.",
    "why_it_matters": "An open dashboard lets users trace four EU Code of Practice risk categories back to 19 benchmarks.",
    "limitations": [
      "The tool is an open evidence interface, not an official EU compliance determination."
    ],
    "importance": 9,
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-24-003/worst-case-scoring-moved-ai-risk-ratings-by-up-to-37-points",
    "json_url": "https://themachinepress.com/story/mp-2026-09-24-003.json",
    "first_published_at": "2026-09-24T09:00:00.000-04:00",
    "modified_at": "2026-09-24T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-09-24-003-001",
        "text": "An open dashboard lets users trace four EU Code of Practice risk categories back to 19 benchmarks.",
        "source_ids": [
          "source-2026-09-24-003"
        ],
        "qualification": "The tool is an open evidence interface, not an official EU compliance determination."
      }
    ],
    "source_ids": [
      "source-2026-09-24-003"
    ],
    "tags": [
      "EU AI Act",
      "systemic risk",
      "benchmarks"
    ],
    "image_url": "https://themachinepress.com/issues/2026-09-24/systemic-risk-laptop-file.webp",
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-09-24-003",
      "title": "arXiv preprint 2609.28335",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2609.28335",
      "canonical_url": "https://arxiv.org/abs/2609.28335",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-09-23T12:13:36.000-04:00",
      "accessed_at": "2026-09-24T08:17:17.000-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-09-24-003-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "Worst-Case Scoring Moved AI Risk Ratings by Up to 37 Points",
    "publisher": "The Machine Press",
    "published_at": "2026-09-24T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-24-003/worst-case-scoring-moved-ai-risk-ratings-by-up-to-37-points"
  }
}