Every action, accounted for.
The Retypd platform is built to work inside regulated operations. Security is how the product was designed, not a layer added later — and we will walk your reviewers through every part of it.
| Deployment | Your cloud or on-premises |
|---|---|
| Model training | Never on customer data |
| Approvals | Human sign-off on irreversible steps |
| SOC 2 Type II | Audit in progress |
Runs where your data already lives.
The same platform, controls and evidence in every model. Most regulated teams run Retypd inside their own cloud account.
| Model | Where it runs | Keys held by | Typical for |
|---|---|---|---|
| Your cloud | Your own AWS, Azure or Google Cloud account, inside your VPC | Yours | Regulated production workloads |
| On-premises | Kubernetes in your data centre, with models you host | Yours | Legacy estates and strict data boundaries |
| Retypd-hosted | A dedicated single-tenant environment operated by Retypd | Yours (BYOK) | Evaluations and pilots |
Your data stays yours.
- Customer data is never used to train models — ours or any provider’s. That commitment is written into every agreement.
- In your cloud or on-premises, data and recordings never leave your environment.
- Sensitive fields are masked at capture, before anything is stored.
- Retention is set per workflow, with deletion at the end of an engagement.
- The Retypd Governance evidence ledger is append-only: records can be exported, never edited.
Agents held to a higher standard than staff.
Every agent gets the scrutiny you give your most privileged employees — and a few controls no employee would tolerate. Policies, approvals, autonomy levels and evidence all live in Retypd Governance.
- 01
Agent identity
Every agent runs under its own named identity with least-privilege access. Credentials sit in the Retypd Connect vault and are never shown to the agent in plain text.
- 02
Access
SSO via SAML or OIDC, and role-based access for who can author, approve and promote a Playbook.
- 03
Approvals
Irreversible steps are held for a named approver, under thresholds written as policy in Retypd Governance.
- 04
Encryption
TLS in transit and AES-256 at rest, with keys you hold in every deployment model.
- 05
Isolation
Single-tenant execution; agent workers are ephemeral and destroyed after every run.
- 06
Secure development
Mandatory code review, signed builds, and static and dependency scanning on every change.
Agents prepare. People decide.
Four commitments that apply to every deployment and every agent.
- I
Human oversight
Autonomy is earned on evidence you define and revoked on drift. Irreversible actions need the approval your policy requires.
- II
Evaluation before release
Every Playbook or model change is replayed against your own historical cases before it reaches production.
- III
Transparency
Agents are identified as agents in every system they touch, and every action can be explained from its evidence in Retypd Governance.
- IV
Contestability
Decisions that affect individuals are made by people. Agents prepare the work; people decide, and can be challenged.
Where things stand, stated precisely.
Ask for the detail behind any line and our security team will walk your reviewers through it.
Request our security overview| SOC 2 Type II | Audit in progress |
|---|---|
| Data processing agreement | Available, including standard contractual clauses |
| Security questionnaires | We complete yours — SIG, CAIQ or your own |
| Model-risk documentation | Prepared per deployment for your second line |
| Vulnerability reports | security@retypd.com — acknowledged within one business day |
Put us through your review.
Our engineers answer questionnaires, join architecture reviews and sit with your second line — before a single agent is deployed.