---
{
  "id": "ASD-RSR-EN-AI-PRIVACY",
  "path": "/en/ai/privacy",
  "title": "Researcher data and access",
  "title_en": "Researcher data and access",
  "category": "Researcher / 研究协作",
  "summary": "Local retrieval; questions and selected public passages are sent only for an explicitly requested model answer.",
  "related": [
    "/researcher",
    "/ai/knowledge-base",
    "/ai/evaluation"
  ],
  "template": "document",
  "version": "1.0.0",
  "updated": "2026-09-27",
  "date_published": "2026-09-27",
  "canonical": false,
  "source_tier": "published",
  "depends_on": [],
  "open_problems": [],
  "equations": [],
  "sources": [
    {
      "label": "ASD published site corpus / Researcher v1.0 release",
      "kind": "website-engineering"
    }
  ],
  "theory_status": "FOUNDATION",
  "evidence_status": "DOCUMENT",
  "locale": "en",
  "translation_path": "/ai/privacy",
  "canonical_url": "https://asd-physics.org/en/ai/privacy",
  "seo_title": "Researcher data and access · ASD Researcher v1.0",
  "metadata_reviewed": "2026-09-27",
  "translation_of": "/ai/privacy",
  "translation_version": "1.0.0"
}
---

## Local retrieval and drafts

The browser downloads the public snapshot and searches locally without sending questions to a model. Questions, temporary access tokens, notes and answers remain in page memory. There is no draft database or automatic publication. You control downloaded exports.

## Requesting a model answer

The interface asks for explicit consent to send the current question, mode, snapshot/page context and server-selected public evidence to the configured OpenAI API. No private-file upload is provided. Do not include information you are unwilling to send. Requests use `store: false`; this does not assert zero provider retention. Provider account settings and terms still apply.

## Access and cost

Model keys stay server-side. Browsers use separate research invitation tokens. The service enforces per-identity frequency, concurrency, length and conservative daily/lifetime cost reservations. Reservations are not invoices; operators must configure price bounds for their chosen model. API billing is separate from ChatGPT subscriptions.

## Records and publication

Quota storage contains only hashed identity, counts, time windows and expiring leases, not questions or answers. The application does not log bodies; infrastructure/provider logging policies still apply. Live evaluation files remain local unless an operator explicitly publishes reviewed copies. Review privacy and draft status before sharing.