Protect data

Governance profiles

A governance profile is the bundle of rules an app follows. Every API key has exactly one, so the protection an app gets depends on which key it uses.

What a profile decides

SettingWhat it does
FrameworksRegulations and standards to enforce, such as HIPAA, GDPR or SOC 2. They become routing limits; when several apply, the strictest wins.
Sensitive dataFor each kind of data (health, personal, card, secrets, government, confidential): Allow, Redact or Block.
GuardrailsPrompt-injection sensitivity, the safety model, which unsafe categories block, denied topics and content limits.
Content captureOff (metadata only), redacted, full or a sample of requests. Captured content is encrypted with your organisation's key.
RetentionHow long audit records are kept, for example six years for HIPAA.
Response cacheAnswer repeat questions from an encrypted cache. Health, card and secret data are never cached by default.
Risk tiersWhat happens to tier 2 and tier 3 risk requests: record, hold for review, or block.
Model allow-listOptionally limit the key to specific models, on top of policy rules.

Open a profile in Governance profiles (opens in a new tab) to see a plain-language "What this profile does" panel that updates as you edit.

Built-in profiles

Every organisation starts with 43 ready-made profiles. Most organisations use them as they are.

CategoryProfiles
HealthcareHIPAA · Clinical; HIPAA · Patient-facing assistant; HIPAA + HITRUST; Healthcare · EU; Clinical research (de-identified); Revenue cycle & claims; Mental & behavioural health; Clinical decision support; Telehealth & virtual care; Life sciences R&D
Financial servicesPCI DSS · Payments; Banking (GLBA + NYDFS); Public company finance (SOX); EU financial services (DORA + GDPR); Insurance claims & underwriting; Fraud & AML; Wealth & investment advisory
GovernmentFedRAMP Moderate; Defence contractors (CMMC / CUI); Law enforcement (CJIS); Citizen services
EducationEducation (FERPA); K-12 schools; Higher education & research
PrivacyGDPR · EU personal data; CCPA / CPRA · California; India DPDP Act
AI governanceEU AI Act · High-risk system; AI management system (NIST AI RMF + ISO 42001); HR & recruiting
General useInternal assistant; Coding agents & developer tools; Customer support assistant; Public website content; Legal & privileged; Maximum protection; Evaluation & sandbox; Retail & e-commerce; Contact centre; Marketing & communications; Security operations; Data analytics & BI; Translation & localisation
  • Secrets are always blocked. Every built-in profile blocks credentials and API keys.
  • Safety model: Automatic. Built-in profiles pick your own on-prem or private-cloud guardrail model. A public one is never chosen, so screening can't send data anywhere the request itself couldn't go. With none configured, the step is skipped.
  • Edit freely. Built-in profiles can be changed; Restore defaults puts one back. Copy and customise in the profile library starts a new profile from any template, and deleted ones can be re-added.

Which profile should I use?

You rarely need to choose by hand: Connect an app (opens in a new tab) asks who uses the app and what data it sees, then recommends one and says why. Some common answers:

AppRecommended profile
Claude Code, Cursor or another coding toolCoding agents & developer tools
Clinical tool used by staffHIPAA · Clinical
Patient or customer-facing health assistantHIPAA · Patient-facing assistant
Customer support botCustomer support assistant
Public website chatPublic website content
Anything that helps decide about people (hiring, credit)HR & recruiting, or EU AI Act · High-risk system for EU residents
Internal assistant with general business dataInternal assistant
Check what a key can actually use
Open API keys (opens in a new tab) and select a key. For each data class, the panel lists which models a request would be routed to (primary and fallbacks), whether values are redacted, or why it would be blocked.