{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-10-08-010",
    "source_story_id": "tmp-story-forget-only-unlearning-memory",
    "edition_id": "mp-2026-10-08-morning-0090",
    "edition_url": "https://themachinepress.com/edition/2026-10-08",
    "position": 10,
    "story_type": "dispatch",
    "section": "safety-security",
    "editorial_classification": "editorial",
    "headline": "Deleting Later May Require Remembering Everything Now",
    "slug": "deleting-later-may-require-remembering-everything-now",
    "dek": "Lower bounds show that forget-only unlearning can need information ordinary training was designed to discard.",
    "summary": "Lower bounds show that forget-only unlearning can need information ordinary training was designed to discard.",
    "body_text": "Forget-only unlearning gives a deletion algorithm the trained model and examples to remove, but no retained data or training trace. The paper shows why that setup cannot always reproduce retraining: different datasets can produce the same model yet require different outputs after the same deletion. It derives accuracy lower bounds for several standard learners and then asks how much training information successful deletion must preserve. For simple threshold learners, arbitrary deletion requests can require memory as large as the full dataset even though ordinary training keeps only a boundary point. The result is theoretical, not a measurement of current large-model unlearning, but it exposes a design conflict: minimal sufficient state for prediction may be insufficient state for future erasure.",
    "why_it_matters": "Lower bounds show that forget-only unlearning can need information ordinary training was designed to discard.",
    "limitations": [
      "Forget-only unlearning gives a deletion algorithm the trained model and examples to remove, but no retained data or training trace.",
      "The paper shows why that setup cannot always reproduce retraining: different datasets can produce the same model yet require different outputs after the same deletion.",
      "For simple threshold learners, arbitrary deletion requests can require memory as large as the full dataset even though ordinary training keeps only a boundary point."
    ],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-08-010/deleting-later-may-require-remembering-everything-now",
    "json_url": "https://themachinepress.com/story/mp-2026-10-08-010.json",
    "first_published_at": "2026-10-08T09:00:00.000-04:00",
    "modified_at": "2026-10-08T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-10-08-010-001",
        "text": "Lower bounds show that forget-only unlearning can need information ordinary training was designed to discard.",
        "source_ids": [
          "source-2026-10-08-010"
        ],
        "qualification": "Forget-only unlearning gives a deletion algorithm the trained model and examples to remove, but no retained data or training trace."
      }
    ],
    "source_ids": [
      "source-2026-10-08-010"
    ],
    "tags": [
      "machine unlearning",
      "data deletion",
      "memorization"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-10-08-010",
      "title": "arXiv preprint 2610.10519",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2610.10519",
      "canonical_url": "https://arxiv.org/abs/2610.10519",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-10-07T13:55:46.000-04:00",
      "accessed_at": "2026-10-08T08:26:48.909-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-10-08-010-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "Deleting Later May Require Remembering Everything Now",
    "publisher": "The Machine Press",
    "published_at": "2026-10-08T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-08-010/deleting-later-may-require-remembering-everything-now"
  }
}