{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-09-08-006",
    "source_story_id": "tmp-story-shadow-query-private-retrieval",
    "edition_id": "mp-2026-09-08-morning-0061",
    "edition_url": "https://themachinepress.com/edition/2026-09-08",
    "position": 6,
    "story_type": "dispatch",
    "section": "safety-security",
    "editorial_classification": "editorial",
    "headline": "The Vector Database Stored Shadow Questions Instead of the Document",
    "slug": "the-vector-database-stored-shadow-questions-instead-of-the-document",
    "dek": "SHAQ decomposed a document into generated queries so stored embeddings revealed less source text while preserving retrieval performance.",
    "summary": "SHAQ decomposed a document into generated queries so stored embeddings revealed less source text while preserving retrieval performance.",
    "body_text": "Embedding-inversion attacks try to reconstruct text from vectors stored for retrieval. SHAQ changes what is stored: a language model generates diverse shadow queries for each document, and the system embeds those queries rather than the original document itself. Across the authors' retrieval datasets, the defense lowered one reported recovery rate to 0.2104, protected up to 19.50 percent more tokens than baseline defenses and reached as much as 0.7967 MAP@10, including utility gains up to 5.53 percent. These are benchmark results for the paper's attack and retrieval settings, not a guarantee against every inversion technique.",
    "why_it_matters": "SHAQ decomposed a document into generated queries so stored embeddings revealed less source text while preserving retrieval performance.",
    "limitations": [
      "These are benchmark results for the paper's attack and retrieval settings, not a guarantee against every inversion technique."
    ],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-08-006/the-vector-database-stored-shadow-questions-instead-of-the-document",
    "json_url": "https://themachinepress.com/story/mp-2026-09-08-006.json",
    "first_published_at": "2026-09-08T09:00:00.000-04:00",
    "modified_at": "2026-09-08T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-09-08-006-001",
        "text": "SHAQ decomposed a document into generated queries so stored embeddings revealed less source text while preserving retrieval performance.",
        "source_ids": [
          "source-2026-09-08-006"
        ],
        "qualification": "These are benchmark results for the paper's attack and retrieval settings, not a guarantee against every inversion technique."
      }
    ],
    "source_ids": [
      "source-2026-09-08-006"
    ],
    "tags": [
      "vector databases",
      "privacy",
      "embedding inversion"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-09-08-006",
      "title": "arXiv preprint 2609.04767",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2609.04767",
      "canonical_url": "https://arxiv.org/abs/2609.04767",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-09-04T01:59:54.000-04:00",
      "accessed_at": "2026-09-08T08:20:00.000-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-09-08-006-001"
      ]
    }
  ],
  "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 Vector Database Stored Shadow Questions Instead of the Document",
    "publisher": "The Machine Press",
    "published_at": "2026-09-08T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-09-08-006/the-vector-database-stored-shadow-questions-instead-of-the-document"
  }
}