Privacy

Your data, your control.

Neuron captures what you ask it to and nothing else. We never train a public model on your data, and you control what's read, who can read what, and how long it's kept.

Zero trainingWorkspace-isolatedRetention you set

What we capture

Only the sources you explicitly connect, and only the channels / repos / labels you select inside each source. Disconnecting a source removes future captures immediately; existing items remain until you delete them (or your retention window expires).

What we never do

  • Train public models on your data. Your messages, docs, and decisions never leave the workspace they came from. Period.
  • Read what you didn't share. If a channel isn't on your include list, Neuron does not request it from Slack's API.
  • Cross-workspace leakage. Knowledge is scoped to its workspace. Cross-workspace search is impossible at the database layer.
  • Sell, share, or rent your data. No data brokers, no ad partners, ever.

Controls you have

Channel-level inclusion

You choose which Slack channels Neuron reads. Defaults to public channels only; private channels and DMs require explicit opt-in. Updates take effect within ~60 seconds.

Per-source retention

Set retention windows per source, e.g., keep Slack for 6 months, Notion forever, Gmail for 30 days. After the window, items are permanently deleted (hard delete, not soft).

Right to delete

Delete any individual knowledge item from the web app. Delete an entire source via Settings → Sources. Delete your entire workspace via Settings → Danger zone. Deletes are propagated to all storage layers (Postgres, vector store, search index) within 24 hours.

Audit log

Every read and write, by humans and by agents, is logged. Audit logs are queryable from Settings → Audit and exportable as CSV/JSON.

How LLMs are used

Neuron uses LLMs for two things: extracting structure from your messages (rules, decisions, processes) and synthesizing answers to queries. We use enterprise contracts with our model providers that prohibit training on inference data and require zero data retention.

Where your data lives

  • Postgres in us-east-1 (primary) and eu-west-1 (EU customers).
  • Vector embeddings in Pinecone, scoped per-workspace by namespace.
  • All data encrypted at rest (AES-256) and in transit (TLS 1.3).
  • Self-host option available on enterprise plans, see Security.

Compliance

  • SOC 2 Type 2, in progress, targeted completion Q1 2027.
  • GDPR-aligned data handling. DPA available on request.
  • CCPA support: California residents can request export or deletion at any time.
  • Sub-processors list maintained in Security.

Last updated 2026-05-31. Questions: privacy@neuron.so.