Attorney Confidentiality
Effective date: August 17, 2026. Last updated: October 16, 2026. Last verified: October 16, 2026. This statement describes how Docket Daily treats attorney search, Workspace, and compliance profile data. It supplements our Privacy Policy.
Docket Daily is built for licensed attorneys who need to monitor legal developments without exposing their research interests or client-specific compliance profiles to staff, advertisers, or AI training pipelines. The Service is exclusively for users who meet the Licensed Attorneys ONLY eligibility in the Terms of Service.
1. Search queries
- In-app: Your dashboard search terms are not stored in our database, not written to application logs, and not associated with your account for analytics or profiling.
- Telemetry: We may count that a search occurred (e.g., for product reliability), but not the text you typed.
- Infrastructure: Search terms are sent in POST request bodies, not URL query strings, so they do not appear in routine Vercel access-log URL fields. Network infrastructure (hosting, CDN, database pooler) may still briefly process request metadata in the ordinary course of transmission; we minimize query exposure and do not intentionally retain search text in those layers.
- AI: Search queries and compliance profile contents are not sent to Anthropic or any other model provider for summarization, suggestions, or training.
2. Compliance profiles and organization data
- Workspace data is protected by PostgreSQL row-level security (RLS)scoped to your organization: a row is visible only when your session is a member of that organization and your role permits the action. Cross-organization probes return empty; they do not reveal another organization's rows.
- Docket Daily staff cannot browse subscriber compliance profiles or Workspace client records through admin tools or service-role shortcuts exposed to the application UI.
- Tier gate: Professional and Enterprise subscriptions are required at both the application layer and the database RLS layer to read or write compliance profile data.
3. AI and third-party use of your inputs
- Corpus summarization and Compliance Suggestions use public, published legal instruments from our editorial corpus—not your search history or profile fields.
- We do not use subscriber search or profile data to train third-party AI models.
Product AI / embeddings subprocessors
| Vendor | Role | Retention / training |
|---|---|---|
| Anthropic, Inc. | Product AI generation (Messages API) | Zero Data Retention: active (effective Oct 16, 2026) — applies to Docket Daily's Anthropic organization and eligible API features. |
Search and relevance embeddings are computed on Docket Daily's own infrastructure — this text never leaves our systems.
Model providers — scope of use. Anthropic (generation): produces the summaries and analysis you read, under no-training commercial terms with Zero Data Retention active for our organization (scope above). Embeddings for Ask/pgvector search and ingest relevance scoring run on Docket Daily's own infrastructure (open-source ONNX model; identifier-scrubbed input). No third-party embeddings vendor receives query or corpus text. Client and matter identifiers are scrubbed server-side before any text reaches Anthropic. Our build pipeline blocks any OpenAI API host or SDK call.
- Ask, Watch Scans compile, and Redline impact drafting scrub identifiers before the model is called. Do not enter client-confidential or nonpublic information.
- Ask: question and answer history is stored encrypted for thirty (30) days, then deleted.
- Document Review: uploaded files are stored in a private bucket and retrieved with signed URLs; findings are encrypted at rest and retained for thirty (30) days (or deleted immediately upon request).
- Client names are encrypted at rest and are never transmitted to any model provider.
- An organization administrator may disable all AI features for the organization.
4. Advertising and privacy signals
- Docket Daily does not sell subscriber data for advertising.
- We honor Global Privacy Control (GPC) and on-site advertising opt-out choices. Signed-in subscribers can opt out account-wide in one click; visitors who are not signed in can opt out this browser. See Your Privacy Choices.
Revision history
October 16, 2026 — Self-hosted embeddings cutover (EM1-c). Search and relevance embeddings moved onto Docket Daily's own infrastructure (open-source ONNX model). Removed OpenAI as an embeddings subprocessor. Anthropic remains the sole AI model provider for generation.
October 16, 2026 — Anthropic Zero Data Retention confirmed active (previously requested/pending). Applies to Docket Daily's Anthropic organization and eligible API features.
September 14, 2026 — Embeddings subprocessor. Named embeddings-only limited-purpose processing (scrubbed input; no generation). Removed the prior embeddings vendor from the current list. Anthropic remains the generation subprocessor.
September 13, 2026 — Embeddings subprocessor. Named Voyage AI (voyage-4) as the product embeddings subprocessor (identifier-scrubbed input; zero-day retention under hosted-API opt-out terms). Anthropic remains the generation subprocessor.
Limitations (honest infrastructure caveat)
No web application can guarantee zero transient processing by hosting providers, DNS, TLS terminators, or database connection poolers. We design to avoid intentional retention and to keep query text out of URL-based access logs. For questions, contact brad@docketdaily.ai.