Your work stays yours.
River is the AI workspace that does not train on your content, does not share it with third parties beyond what's needed to respond, and supports on-prem deployment for anyone who needs full control.
Free to start · No credit card required
Most AI tools quietly use your conversations to train their next model. River doesn't. Your drafts, your spaces, your messages stay yours — not part of someone else's training corpus.
For teams that need stronger guarantees, the GalBox on-prem option runs the entire River stack on hardware you own, with no outbound data flow at all.
What 'private' actually means here
No training on your content
River does not train its models on your documents, your chat history, or your Space content. Ever.
Encrypted everywhere
TLS in transit, AES-256 at rest, per-org isolation. The standard, applied consistently.
On-prem GalBox option
Run the entire River stack on hardware in your office. Air-gapped if required. No outbound calls.
Tight access controls
Per-Space sharing, per-document permissions, full audit log of who saw what when.
Minimal third-party exposure
Inference goes only to the model provider needed for the response — no analytics piping, no observability uploads of your prose.
Export and leave
Plain-text export of everything, anytime. No data lock-in, no exit tax.
Three deployment shapes
- 1
Cloud (default)
River's hosted multi-tenant cloud with strict isolation. Free to start, scales to teams of any size.
- 2
Cloud private tenant
Single-tenant deployment for organisations that need a dedicated environment. Encrypted with your own KMS keys.
- 3
On-prem GalBox
The full River stack on a box in your office. Air-gapped, no outbound traffic, runs locally.
Frequently asked questions
How do I know you don't train on my data?
Our Terms commit to it explicitly, our enterprise contracts add a separate DPA, and the on-prem deployment removes the question entirely — our servers never see your content.
What about the LLM provider — do they train?
We route to providers that offer no-training inference by contract (Anthropic, OpenAI enterprise, Google enterprise). For maximum guarantees, the on-prem option uses open-weight models running locally.
Can I see what's been shared?
Yes. Every Space has an audit log of who accessed what when. Per-document permissions let you tighten sharing for sensitive work.
What is GalBox?
GalBox is River's on-prem appliance — hardware you buy that runs the full River stack locally. No internet required. Used by family offices, defence contractors, healthcare systems, and labs with sensitive research.