Sapphire Legal
Security & data sovereignty

Not a policy. An architecture you can check.

Every practice owes its clients confidentiality, and most software asks you to take separation on trust — one database, one shared table, a column to tell the tenants apart. Sapphire Legal is architected around real per-tenant isolation and a private AI instead. It matters most sharply when two of your clients compete, which is a fractional GC's ordinary week, and it is the same architecture under a solo practice, a law firm and a regulated organization.

Quick answer
How does Sapphire Legal keep client data isolated?
Sapphire Legal gives each client firm its own database, its own document storage, and a private legal AI that runs inside our own AWS account — not a public chatbot. Each client's data lives in a different place from every other client, not in a shared database with a tenant ID column. The AI that drafts contracts and analyzes calls is served privately through Amazon Bedrock inside that account: prompts and responses are never used to train a model — Bedrock's terms do not permit it — and nothing is retained once the request completes. Privileged client call transcripts are locked behind an attorney review step that's recorded in an audit log. That is what a duty of confidentiality actually requires, and it is what holds up in the hardest case — when two of your clients are competitors.
Three pillars

Isolation is the foundation, not a feature.

Most multi-tenant software puts all clients in one database and relies on code to keep them apart. That's fine for a CRM. It is not fine for privileged client files. Here's how Sapphire Legal is different.

Per-tenant database

Every client firm gets its own database — not a shared table with a tenant column, but a separate database entirely. Data cannot leak between clients because it doesn't live in the same place to begin with.

Per-tenant storage

Each client's documents are stored in an isolated location with encryption at rest. Optional dedicated storage buckets are available for clients whose compliance team wants physical separation.

Private legal AI in AWS

The AI that drafts your documents and analyzes your calls runs privately inside our own AWS account, in US regions — models we host ourselves alongside Claude models served privately through Amazon Bedrock. It's not a public chatbot. Your client data is never used to train a model, and nothing is retained once a request completes.

Sapphire Clarity

Privileged calls stay privileged — and we can prove it.

Upload a client-call recording to Sapphire Clarity and it becomes searchable, structured legal intelligence. When the call is marked attorney-client privileged, the resulting transcript is locked behind an explicit review step. An attorney has to confirm they've reviewed the material before anyone else on the team can open it — and that review is logged with name, timestamp, and justification.

If privilege is ever challenged, you have a defensible audit trail showing exactly who accessed what and when.

Sapphire Clarity legal risk flags — four open flags from a witness interview including a high-severity sincerity drop at 0:47 flagged as possible evasion, a financial transaction admission, a prior conviction disclosure classified as a credibility signal, and a chain-of-custody issue on acknowledged property
  • Attorney review required before transcript access
  • Audit log with reviewer name + timestamp
  • Privileged flag persists across exports
  • Access can be revoked at any time
  • Conflict check runs automatically on intake
FAQ

Security questions, answered plainly.

Does Sapphire Legal train AI models on my client data?

No. The models are served privately through Amazon Bedrock, inside Sapphire Legal’s own AWS account. Your prompts and documents are never used to train a model — Bedrock’s terms do not permit training on customer content — they are not retained once the request completes, and they are not pooled with any other customer’s data. AWS serves the model; the company that built it never receives your material. It is used to answer your questions and nothing else.

Is my data shared with OpenAI, Anthropic or any other AI provider?

Nothing is sent to OpenAI, and nothing is sent to a public chatbot. The models that draft documents, analyze calls and answer questions are Claude models served through Amazon Bedrock, which runs inside Sapphire Legal’s own AWS account. AWS serves the model there; Anthropic, which builds it, never receives your prompts or documents. Bedrock’s terms do not permit training on customer content, and our IAM policy restricts the path to US-only model profiles, so your material cannot be processed outside the United States.

What does Sapphire Legal log about my use of the AI?

Usage, not content. Every model call is metered per firm, per feature and per model — tokens used, what it cost, how long it took, whether it succeeded. The prompt itself and the model’s response are not written to that record, and Bedrock’s own invocation logging is not switched on. So Sapphire Legal can tell you how much the AI was used on a matter, and cannot tell you what it said.

How is my firm’s data separated from other firms?

Every client firm gets its own PostgreSQL database and its own document storage — a physically separate database, not a shared table with a tenant identifier column. Data cannot leak between firms because it does not live in the same place to begin with. This is the distinction that matters the moment the files are privileged, and most sharply when two clients of the same practice compete.

Is Sapphire Legal confidential enough for privileged material?

Privileged client call transcripts sit behind an attorney review step, and that review is written to an audit log. Combined with per-tenant isolation and a private in-account AI, that means privileged material is never exposed to another tenant, is never used to train a model, and is not retained once a request completes. Access to a privileged transcript is recorded. The audit trail is designed to be something you can show a client or a regulator.

Where does my data physically live?

In Sapphire Legal’s AWS infrastructure, in the region for your market. Each firm’s database and document storage are separate resources within it, so “where does my data live” has a specific answer per firm rather than a shared one. Current regional detail for your market is available on request during onboarding.

Who at Sapphire Legal can see my documents?

Access to customer data is restricted and logged. Routine product operation does not require anyone at Sapphire Legal to read your documents, and support access to a tenant is deliberate and recorded rather than ambient. Because each firm is a separate database, there is no internal view that spans customers by default.

What happens to my data if I cancel?

Your firm’s data remains yours. Export is available, and the retention and deletion handling for your tenant is set out in the Terms and the Privacy Policy rather than left to practice. Because your data is its own database, removing a tenant removes that data rather than filtering it out of a shared store.

Does the AI show where an answer came from?

Yes. Ask Sapphire answers only from your firm’s own records, and every statement carries a numbered source that opens the document, register or report it rests on. If the answer is not in your records, it says so rather than filling the gap. Actions it prepares — tasks, deadlines, drafts, filed email — are cards you approve, and each approval is audit-logged.

Need the technical deep-dive?

If your clients' security team wants the full architecture diagram — database isolation, token verification, audit logging, the whole thing — we have it ready. Book a demo and we'll send it over before the call.

Pricing →