Your doctrine, signals, and transcripts.
Here is exactly how we protect them.

Every Keel Workspace is created in exactly one region. The region is locked at creation; data, transcripts, embeddings, and audit logs do not cross regional boundaries. v1 ships in us_east. EU and APAC regions are on the v2 roadmap; if you need data residency in those regions before then, contact us before provisioning so we can plan accordingly.

• Region pinning
  Locked at Workspace creation. One-way door.
• Encryption at rest
  All data classes. Per-region keys.
• Encryption in transit
  TLS 1.2 or higher on all connections.
• Transcript storage
  Cloudflare R2, region-pinned buckets.
• Embeddings
  Region-scoped vector store.
• Audit logs
  Region-scoped, append-only.

Customer-managed encryption keys (BYOK) are on the v1.2 roadmap for enterprise tenants. Until then, encryption keys are managed by Keel and never leave the customer's region.

Two categories of access exist: your team, and authorized Keel staff. Both are scoped, both are audited, and both appear in your audit log.

Your team

Hierarchical roles: owner, admin, publisher, editor, member. Each role is a strict superset of the role below it. Coaching workspace access is gated by manager-of relationship, not by role, and is explicitly framed as not for performance review. SMEs and partners who do not need a seat are modeled as Persons, with optional time-boxed Guest Invites for deal-specific access.

Authorized Keel staff

Phase 1 design partners receive audit-only support access by default (per the DPA addendum signed at onboarding). This grants authorized Keel staff read access to your audit log, Blueprint history, Vera interaction summaries (timestamps and token usage, not full payloads), and integration health. It does not grant access to deal contents, contact details, transcripts, or Vera prompt and response payloads. It does not grant the ability to act on your team's behalf.

Every Keel staff read produces an entry in your audit log with actor type, staff member, reason text, and target. You can see exactly when and why we read anything. Audit access can be disabled by your admin at any time. Standard plan customers have audit access default-off and opt in if and when they want our support team to have it.

• SOC 2 Type I

  In progress

  Phase 1. Report available on request once issued.

• SOC 2 Type II

  Roadmap

  Begins observation period after Type I issuance.

• External penetration test

  Scheduled

  Annual cadence. Summary available on request.

• GDPR readiness

  In place

  DPA available for signature. EU sub-processor flow documented.

• CCPA readiness

  In place

  See Privacy Policy for California-specific rights.

• UK GDPR readiness

  In place

  DPA covers UK data subjects.

Phase 1 design partners onboard before SOC 2 Type I is issued. We document this explicitly in the design partner agreement and provide a Type I roadmap with target dates as part of the security review package. Customers who require Type II in hand before signature should reach out to discuss timing.

See the Security FAQ ->

Vera is the AI agent inside Keel. She runs on LLM providers your team configures. Three principles govern how AI is used.

• Bring your own provider
  Use your own OpenAI, Anthropic, Google, or Azure OpenAI key. Or use Unfold AI as the default fallback.
• No training on your data
  Provider contracts disable training where supported. Sub-processor list specifies the exact terms per provider.
• Pattern Library is opt-in
  Default off. Aggregates are de-identified. Deal contents, contact details, and transcripts never contribute.

Token budget controls

Admins set a Workspace-level monthly token cap with alerts at 50%, 80%, and 100%. Per-Blueprint sub-budgets are available for multi-practice firms. At cap, non-critical operations degrade gracefully (soft cap) or pause (hard cap, opt-in). No surprise invoices.

Provider error handling

Provider errors surface as typed errors with explicit remediation:models_not_configured, provider_unauthorized, provider_quota_exceeded, provider_unavailable, provider_request_invalid. Where the error means your BYO setup is blocked, an in-product CTA offers a one-click switch to the Unfold AI fallback so your team is not stuck. Org admins can suppress this CTA where partner contracts require it.

Keel uses a small number of sub-processors to deliver the product. The full list with regions and purposes lives on the sub-processors page; the headline categories are below.

• Cloud hosting
  Cloud Run (compute), Cloudflare (CDN, R2 storage)
• Database
  Managed Postgres in customer's region
• LLM providers (BYO + default)
  OpenAI, Anthropic, Google, Azure OpenAI, Unfold AI
• Email delivery
  Transactional email service (notifications, alerts, magic links)
• Observability
  Error tracking and uptime monitoring
• Identity (v1.1)
  SSO and SCIM providers customer chooses

See the full sub-processor list ->

• Policy
  Terms of Service

  Terms governing use of the website and platform; relationship to the MSA for paid customers.

• Policy
  Privacy Policy

  What we collect, why, how we use it, retention, and your rights.

• Agreement
  Data Processing Agreement (DPA)

  Standard DPA with EU SCCs and UK addendum. Available for signature.

• Reference
  Security FAQ

  Hosting, encryption, access controls, incident response, backup.

• Policy
  Vulnerability Disclosure

  Scope, reporting channel, safe harbor, response SLA.

• Reference
  Sub-processor list

  Every system Keel uses to deliver the product, with regions and purposes.

For security review packages, vendor questionnaires, and scheduling time with our team, write to hello@keelflow.ai. For vulnerability reports, write to security@keelflow.ai (see our disclosure policy for scope and safe-harbor terms). For data subject requests under GDPR, CCPA, or other applicable law, write to privacy@keelflow.ai.

This page is updated as our certifications progress and as our sub-processor list changes. Substantive updates are announced in-product and by email to designated security contacts.