Trust Center
Security and privacy, stated plainly.
What we have in place, what we are building, and what we do not yet hold. SOC 2 and ISO/IEC 27001:2022 are in progress, and we say so. Updated 4 November 2026.
- controls in place
- 50
- approved policies and documents
- 62
- subprocessors listed
- 32
- last updated
- 4 Nov 2026
Every control on this page works for customers today. What we are still building or rolling out, and what remains for each, is in our security questionnaire and system description, which we share under a confidentiality agreement: request documents. We do not hold a SOC 2 report or an ISO/IEC 27001 certificate yet.
Compliances
Where each framework stands.
A status, never a badge. Each card says what we hold, what we are doing, and what is still ahead.
SOC 2
In progressWe are preparing for a SOC 2 Type I audit and will follow it with a Type II observation period. Neither is complete, and we do not hold a SOC 2 report.
The AICPA's framework for how a service organisation protects customer data: security, availability, processing integrity, confidentiality and privacy.
ISO/IEC 27001:2022
In progressWe are building our information security management system to ISO/IEC 27001:2022. The certification audit has not been completed, and we do not hold a certificate.
The international standard for running an information security management system.
DPDP Act 2023 (India)
Aligned controls, program in placeWe run a privacy program and are aligning our controls with India's Digital Personal Data Protection Act and its Rules: a privacy notice, a grievance route, processor terms for customers, and tools for rights requests and retention that are still being built.
India's data protection law. It has no certification scheme, so this describes our program, not a badge.
GDPR
DPA on requestWhere we process EU personal data for a customer, we act as its processor. Processor terms (a data processing agreement) are available on request.
The European Union's data protection regulation.
Controls
What is in place today.
One sentence for each control that works for customers today. The framework references under each one show which criteria it is meant to serve; they do not say an audit has found it effective. What we are still building is in the documents we share on request.
Data security
How customer data is protected, kept to what is needed, and removed on request.12 controls in place12 controls in place
- Implemented
Encryption at rest
Our database, object-storage and hosting providers encrypt stored data at rest.
SOC 2 CC6.1, C1.1 · ISO/IEC 27001 A.8.24 · DPDP Rule 6(1)(a)
- Implemented
Encryption of call content inside our application
Since 4 Nov 2026, the transcript, summary, analysis, extracted answers and tool-call logs in each new business call's record are encrypted in our primary database with AES-256-GCM, under one platform-wide key protected by AWS KMS. Call recordings, calls stored before that date, and the copies our services keep outside the call record are not covered by this layer and rely on our providers' encryption at rest.
SOC 2 CC6.1, C1.1 · ISO/IEC 27001 A.8.24, A.8.11 · DPDP Rule 6(1)(a)
- Implemented
Customer credentials in a secrets store
Carrier and in-app push credentials that customers connect are kept in AWS Secrets Manager, not in our databases, and push credentials are write-only: once saved, they are never shown again, not even to our staff.
SOC 2 CC6.1 · ISO/IEC 27001 A.5.17, A.8.24 · DPDP Rule 6(1)(a)
- Implemented
Transcript redaction
You can turn on redaction per agent so that identifiers such as phone numbers, Aadhaar, PAN, card numbers and email addresses are removed from transcripts before they are stored.
SOC 2 C1.1, P3.1 · ISO/IEC 27001 A.8.11, A.5.34 · DPDP Rule 6(1)(a)
- Implemented
Agent memory controls
Agent memory is off until you turn it on, what is saved is redacted first, it is keyed by a one-way hash of the user id, and one person's memory can be erased.
SOC 2 P4.2, C1.2 · ISO/IEC 27001 A.8.10, A.5.34 · DPDP s.8(7)
- Implemented
Masking on the device
The in-app SDK masks PAN, Aadhaar, card numbers, phone numbers and email addresses in the screen context, variables and user traits it sends on its own, before they leave the device.
SOC 2 C1.1, CC6.7 · ISO/IEC 27001 A.8.11 · DPDP Rule 6(1)(a)
- Implemented
No training on your data
Graine does not use customer data to train or fine-tune general AI models, and a custom voice is made only from recordings the customer gives us for it; speech and language vendors apply their own terms, and some run model-improvement programs.
SOC 2 P4.1 · ISO/IEC 27001 A.5.34
- Implemented
Private voice clones
A cloned voice is visible only to the member who made it, and no one else can put it on an agent in Graine.
SOC 2 P2.1, CC6.1 · ISO/IEC 27001 A.5.34, A.5.15 · DPDP s.6
- Implemented
Data requests by person
An Admin can find one person's data across calls, transcripts, recordings and contacts, export it, correct it or erase it, and each request is tracked against the dates the law sets. An erasure covers our main database and recording storage; the copy of a call kept in a separate working store is not erased by it.
SOC 2 P5.1, P5.2 · ISO/IEC 27001 A.5.34 · DPDP s.11, s.12
- Implemented
Retention controls
You choose how long recordings, transcripts and campaign contact details are kept; past that they are removed automatically, while the record of each call and its charge stay. The copy of a call kept in a separate working store is not removed by this.
SOC 2 P4.2, C1.2 · ISO/IEC 27001 A.8.10, A.5.33 · DPDP s.8(7), Rule 8
- Implemented
Workspace offboarding
When a customer leaves, their workspace's data is exported and then deleted, with a certificate of what was removed, what the law requires us to keep and what the deletion doesn't reach. The copy of each call kept in a separate working store is not deleted by this.
SOC 2 C1.2, CC6.5 · ISO/IEC 27001 A.8.10, A.5.11 · DPDP s.8(7)
- Implemented
Replicated primary database
Our primary database is a managed three-node replica set run by our database provider.
SOC 2 A1.2 · ISO/IEC 27001 A.8.14, A.8.13 · DPDP Rule 6(1)(d)
Network security
How the platform is hosted, connected and watched.8 controls in place8 controls in place
- Implemented
Production hosting
The call platform runs on AWS in Mumbai, India. The subprocessor list below says where every other kind of data is processed.
SOC 2 CC6.4, A1.2 · ISO/IEC 27001 A.5.23, A.8.14
- Implemented
Encryption in transit
Our website, dashboard, API and call-streaming sockets use TLS 1.2 or 1.3 (older versions are refused), and browsers are told to use HTTPS only.
SOC 2 CC6.1, CC6.7 · ISO/IEC 27001 A.8.24, A.5.14, A.8.20 · DPDP Rule 6(1)(a)
- Implemented
Internal routes closed to the internet
Our load balancer blocks routes meant only for our own services, and accepts phone-number lookups only from our own servers.
SOC 2 CC6.6 · ISO/IEC 27001 A.8.20, A.8.22 · DPDP Rule 6(1)(b)
- Implemented
Service-to-service authentication
Our services prove themselves to each other with a shared secret compared in constant time; where none is configured, a route refuses every request instead of opening up.
SOC 2 CC6.1, CC6.6 · ISO/IEC 27001 A.8.5, A.8.20 · DPDP Rule 6(1)(b)
- Implemented
Rate limits
The public API is rate-limited per organisation and shows your limit in its response headers; the SDK's session routes and our public forms are rate-limited too, and our sign-in provider limits sign-in attempts.
SOC 2 CC6.1, CC7.2 · ISO/IEC 27001 A.8.6, A.8.20 · DPDP Rule 6(1)(g)
- Implemented
Browser security headers
The dashboard and the API tell browsers to use HTTPS only (HSTS), not to sniff content types and not to let other sites frame them; a content-security policy is running in report-only mode.
SOC 2 CC6.6, CC6.7 · ISO/IEC 27001 A.8.20, A.8.26 · DPDP Rule 6(1)(g)
- Implemented
Outbound request protection
Webhook deliveries and agent API tools refuse private and internal network addresses.
SOC 2 CC6.6 · ISO/IEC 27001 A.8.20, A.8.26
- Implemented
Health monitoring
Our services expose health checks and metrics, and send heartbeats to an external uptime monitor.
SOC 2 CC7.2, A1.1 · ISO/IEC 27001 A.8.16, A.8.6 · DPDP Rule 6(1)(c)
Application security
What you control in your workspace, and how the product itself is built and changed.23 controls in place23 controls in place
- Implemented
Roles
Every workspace member is an Admin, a Member or View only, and Admins can create custom roles that hold exactly the permissions they choose; a role is read live, so a change applies without a sign-out.
SOC 2 CC6.1, CC6.3 · ISO/IEC 27001 A.5.15, A.5.18, A.8.2 · DPDP Rule 6(1)(b)
- Implemented
Per-agent access
A person can be limited to chosen agents, and anything outside their access is hidden from them.
SOC 2 CC6.3 · ISO/IEC 27001 A.5.15, A.8.3 · DPDP Rule 6(1)(b)
- Implemented
Maker–checker approvals
A Member's go-live, change to a live agent or campaign launch can wait for another person's approval, and nobody can approve their own request.
SOC 2 CC6.3, CC8.1 · ISO/IEC 27001 A.5.3, A.8.32
- Implemented
A passing test run before go-live
An Admin can require an agent's latest test run to pass before the agent goes live, whether it goes live from the dashboard or through the API.
SOC 2 CC8.1 · ISO/IEC 27001 A.8.29, A.8.32
- Implemented
Workspace isolation
Every request acts only on the caller's own workspace, another workspace's data answers "not found", and an automated test sends a foreign id to every route that takes one.
SOC 2 CC6.1, CC6.3 · ISO/IEC 27001 A.8.3, A.5.15 · DPDP Rule 6(1)(b)
- Implemented
Sub-account isolation
Each sub-account is a separate workspace with its own data and people, and only its parent organisation can manage it.
SOC 2 CC6.1 · ISO/IEC 27001 A.5.15, A.8.3
- Implemented
Scoped API keys
API keys carry scopes, and a key a Member creates can never do more than a Member can.
SOC 2 CC6.1, CC6.2 · ISO/IEC 27001 A.5.16, A.5.17, A.8.2 · DPDP Rule 6(1)(b)
- Implemented
Two-step sign-in
A password sign-in asks for a one-time code sent to the member's email before the session starts.
SOC 2 CC6.1 · ISO/IEC 27001 A.8.5 · DPDP Rule 6(1)(b)
- Implemented
Access reviews
Admins review who has access (roles, agents, sub-accounts and API keys) and record that the review was done.
SOC 2 CC6.2, CC6.3 · ISO/IEC 27001 A.5.18 · DPDP Rule 6(1)(b)
- Implemented
Verifiable audit trail
Changes in your workspace are written to an append-only, hash-chained audit trail that you can verify yourself, in the dashboard or through the API.
SOC 2 CC7.2, CC4.1 · ISO/IEC 27001 A.8.15 · DPDP Rule 6(1)(c)
- Implemented
Complete audit export
A complete export of your audit trail, with a check that its chain is unbroken, for your own archive or SIEM; the dashboard also downloads the Activity log with its proof.
SOC 2 CC7.2 · ISO/IEC 27001 A.8.15, A.5.33 · DPDP Rule 6(1)(c), Rule 6(1)(e)
- Implemented
Security log
A log shows every security check each of your agents makes, with its result and reason, and every change to an agent's security settings with the member who made it; rows are kept for a year.
SOC 2 CC7.2 · ISO/IEC 27001 A.8.15, A.8.16 · DPDP Rule 6(1)(c)
- Implemented
Signed webhooks
Webhooks we send are signed with HMAC-SHA256 over a timestamp and the exact bytes sent, so your systems can check they came from us.
SOC 2 CC6.7 · ISO/IEC 27001 A.8.24, A.5.14
- Implemented
Allow-listed SDK keys
Publishable keys for the in-app SDK and the web widget work only from the websites and apps you allow, and are rate-limited.
SOC 2 CC6.1, CC6.6 · ISO/IEC 27001 A.5.15, A.8.5 · DPDP Rule 6(1)(b)
- Implemented
Verified apps and verified users
For an in-app agent you can require proof of the app (Apple App Attest or Google Play Integrity) and proof of the user (a signed identity or your own sign-in); each is off until you turn it on, with a report-only stage first.
SOC 2 CC6.1, CC6.2 · ISO/IEC 27001 A.5.16, A.8.5 · DPDP Rule 6(1)(b)
- Implemented
Consent for WhatsApp calls
A WhatsApp call goes out only when a consent from the person being called is on record.
SOC 2 P2.1 · ISO/IEC 27001 A.5.34 · DPDP s.6
- Implemented
Privacy notice
Our privacy policy says what we collect, why, who processes it, how long we keep it and how to use your rights.
SOC 2 P1.1, CC2.3 · ISO/IEC 27001 A.5.34 · DPDP s.5, Rule 3
- Implemented
Grievance and rights requests
Anyone can raise a grievance or a data-rights request on our website or by email, gets a ticket number, and gets a reply within 30 days; we never take longer than 90.
SOC 2 P8.1, P5.1 · ISO/IEC 27001 A.5.34 · DPDP s.8(10), s.13, Rule 14
- Implemented
A permission on every route
Every route in our API and dashboard declares the permission it needs, and an automated test fails the build for any route that doesn't.
SOC 2 CC6.1, CC8.1 · ISO/IEC 27001 A.8.26, A.8.29
- Implemented
Isolation and audit tests
Automated suites check workspace isolation, the audit chain, webhook signing and fail-closed internal routes.
SOC 2 CC4.1, CC8.1 · ISO/IEC 27001 A.8.29, A.8.25
- Implemented
Tests in the build pipeline
Every service's test suite runs in our build pipeline on each change, and a failing suite marks the change as failed.
SOC 2 CC8.1 · ISO/IEC 27001 A.8.25, A.8.29, A.8.32
- Implemented
Dependency and secret scanning
Scans for vulnerable dependencies, committed secrets, code weaknesses and misconfiguration run on every change and every week, and report what they find to our engineers.
SOC 2 CC7.1, CC8.1 · ISO/IEC 27001 A.8.8, A.8.29
Corporate security
The program around the product: policies, people, vendors and how we respond.7 controls in place7 controls in place
- Implemented
Subprocessor inventory
We keep a list of every subprocessor that handles customer data, with what each one processes and where, and publish it on this page.
SOC 2 CC9.2 · ISO/IEC 27001 A.5.19, A.5.21 · DPDP s.8(2)
- Implemented
Published data locations
We publish where each subprocessor processes data. A location we have not confirmed shows "On request" instead of a guess.
SOC 2 CC9.2 · ISO/IEC 27001 A.5.19, A.5.23 · DPDP s.16, Rule 15
- Implemented
Processor terms
Customers can ask for our data processing agreement, which sets out how we process personal data on their behalf.
SOC 2 P6.1, CC9.2 · ISO/IEC 27001 A.5.19, A.5.34 · DPDP s.8(2), Rule 6(1)(f)
- Implemented
Incident register
Security incidents are kept in a register with clocks for telling customers and regulators on time.
SOC 2 CC7.3, CC7.4 · ISO/IEC 27001 A.5.24, A.5.26 · DPDP s.8(6), Rule 7
- Implemented
Staff access logging
Each access by Graine staff to a customer workspace is recorded in that workspace's audit trail, with who acted.
SOC 2 CC6.2, CC7.2 · ISO/IEC 27001 A.8.15, A.5.18 · DPDP Rule 6(1)(c)
- Implemented
Vulnerability disclosure
Anyone can report a security issue to us; the contact is published on this page and in our security.txt file.
SOC 2 CC2.3, CC7.1 · ISO/IEC 27001 A.8.8, A.6.8
- Implemented
Internal security review
In October 2026 we reviewed the security of in-app chat and calls against ISO/IEC 27001, SOC 2 and DPDP controls, and fixed in code what the review found that code could fix.
SOC 2 CC4.1, CC3.2 · ISO/IEC 27001 A.5.35, A.8.8
Resources
Policies and documents, open and on request.
Open documents are public once approved. The rest are shared under a confidentiality agreement: you verify your email, accept the agreement, and our team approves each document. A document marked Draft is not yet approved, and the viewer says so.
Privacy Notice
Open to everyone
DPDP s.5 · DPDP Rule 3 · DPDP s.6 · DPDP s.7
Your Data Rights: how to make a request
Open to everyone
DPDP s.11 · DPDP s.12 · DPDP s.13 · DPDP s.14
Security Overview
Open to everyone
SOC 2 CC2.3 · SOC 2 CC6.1 · SOC 2 CC6.6 · SOC 2 CC6.7
Sub-processor List
Open to everyone
SOC 2 CC9.2 · SOC 2 P6.4 · ISO 27001 A.5.19 · ISO 27001 A.5.20
Incident and Breach Notification Commitments
Open to everyone
SOC 2 CC2.3 · SOC 2 CC7.3 · SOC 2 CC7.4 · SOC 2 CC7.5
Business Continuity Overview
Open to everyone
SOC 2 A1.2 · SOC 2 A1.3 · SOC 2 CC7.5 · SOC 2 CC9.1
Data Retention and Deletion
Open to everyone
SOC 2 C1.2 · SOC 2 P4.2 · SOC 2 P4.3 · SOC 2 P5.1
Shared Responsibility Matrix
Open to everyone
SOC 2 CC2.3 · SOC 2 CC6.1 · SOC 2 CC9.2 · ISO 27001 A.5.2
Responsible AI Statement
Open to everyone
EU AI Act Art. 5 · EU AI Act Art. 50 · ISO 42001 A.2.2 · ISO 42001 A.3.2
Data Processing Agreement: summary
Open to everyone
DPDP s.8(2) · DPDP Rule 6(1)(f) · DPDP Rule 7(1) · DPDP Rule 7(2)
Data Processing AddendumDraftCustomers
DPDP s.8(2) · DPDP Rule 6(1)(f) · DPDP Rule 7 · DPDP Rule 8(3)
System Description
SOC 2 CC1.1 · SOC 2 CC2.1 · SOC 2 CC2.3 · SOC 2 CC3.2
Security Questionnaire — Standard Answers
SOC 2 CC1.1 · SOC 2 CC2.3 · SOC 2 CC3.2 · SOC 2 CC6.1
Incident Response Plan
SOC 2 CC2.3 · SOC 2 CC7.3 · SOC 2 CC7.4 · SOC 2 CC7.5
Business Continuity and Disaster Recovery Plan
SOC 2 A1.2 · SOC 2 A1.3 · SOC 2 CC7.5 · SOC 2 CC9.1
Statement of Applicability (ISO/IEC 27001:2022)
ISO 27001 cl. 4.3 · ISO 27001 cl. 6.1.2 · ISO 27001 cl. 6.1.3 · ISO 27001 cl. 8.3
ISMS Scope and Roles
ISO 27001 cl. 4.1 · ISO 27001 cl. 4.2 · ISO 27001 cl. 4.3 · ISO 27001 cl. 4.4
Records of Processing Activities
GDPR Art. 30 · GDPR Art. 28 · GDPR Art. 9 · GDPR Art. 6
Architecture and Data Flow
SOC 2 CC6.1 · SOC 2 CC6.6 · SOC 2 CC6.7 · SOC 2 CC7.1
Penetration Test Status Letter
SOC 2 CC4.1 · SOC 2 CC7.1 · ISO 27001 A.5.35 · ISO 27001 A.8.8
Deployment models and data residency
Penetration test report
Not yet performed. An independent test is planned.
PlannedSOC 2 Type I report
After the audit is completed.
PlannedISO/IEC 27001:2022 certificate
After the certification audit.
PlannedGrievance Redressal Policy
Open to everyone
DPDP s.8(10) · DPDP s.13 · DPDP Rule 14(3) · DPDP Rule 9
Vulnerability Disclosure Policy
Open to everyone
SOC 2 CC2.3 · SOC 2 CC7.1 · SOC 2 CC7.3 · ISO 27001 A.5.25
Acceptable Use Policy
Open to everyone
SOC 2 CC1.1 · SOC 2 CC2.3 · SOC 2 CC6.1 · ISO 27001 A.5.10
Cookie Policy
Open to everyone
DPDP s.5 · DPDP s.6 · DPDP s.6(4) · GDPR Art. 7
Code of Business Conduct
Open to everyone
SOC 2 CC1.1 · SOC 2 CC1.4 · SOC 2 CC1.5 · SOC 2 CC2.2
Information Security Policy
SOC 2 CC1.1 · SOC 2 CC1.2 · SOC 2 CC2.1 · SOC 2 CC2.2
Organisation of Information Security Policy
SOC 2 CC1.2 · SOC 2 CC1.3 · SOC 2 CC1.5 · SOC 2 CC2.2
Compliance Policy
SOC 2 CC1.1 · SOC 2 CC2.3 · SOC 2 CC3.1 · SOC 2 CC4.1
Compliance Procedure
SOC 2 CC2.3 · SOC 2 CC4.1 · SOC 2 CC4.2 · SOC 2 CC5.3
Risk Assessment and Management Policy
SOC 2 CC3.1 · SOC 2 CC3.2 · SOC 2 CC3.3 · SOC 2 CC3.4
HR Security Policy
SOC 2 CC1.1 · SOC 2 CC1.3 · SOC 2 CC1.4 · SOC 2 CC1.5
HR Security Procedure
SOC 2 CC1.4 · SOC 2 CC6.2 · SOC 2 CC6.3 · ISO 27001 A.5.11
Staff Acceptable Use Policy
SOC 2 CC1.1 · SOC 2 CC2.2 · SOC 2 CC6.1 · SOC 2 CC6.8
AI Use Policy
SOC 2 CC1.1 · SOC 2 CC3.2 · SOC 2 CC9.2 · SOC 2 P1.1
Access Control Policy
SOC 2 CC6.1 · SOC 2 CC6.2 · SOC 2 CC6.3 · SOC 2 CC6.6
Access Control Procedure
SOC 2 CC6.1 · SOC 2 CC6.2 · SOC 2 CC6.3 · ISO 27001 A.5.15
Asset Management Policy
SOC 2 CC6.1 · SOC 2 CC6.5 · SOC 2 CC3.2 · ISO 27001 A.5.9
Asset Management Procedure
SOC 2 CC6.1 · SOC 2 CC6.5 · ISO 27001 A.5.9 · ISO 27001 A.5.11
Endpoint Security Policy
SOC 2 CC6.1 · SOC 2 CC6.7 · SOC 2 CC6.8 · SOC 2 CC7.1
Media Disposal Policy
SOC 2 CC6.5 · SOC 2 C1.2 · SOC 2 P4.3 · ISO 27001 A.7.10
Physical and Environmental Security Policy
SOC 2 CC6.4 · SOC 2 A1.2 · ISO 27001 A.7.1 · ISO 27001 A.7.2
Physical and Environmental Security Procedure
SOC 2 CC6.4 · ISO 27001 A.7.1 · ISO 27001 A.7.2 · ISO 27001 A.7.3
Data Classification Policy
SOC 2 C1.1 · SOC 2 CC6.1 · SOC 2 CC6.7 · ISO 27001 A.5.10
Operation Security Policy
SOC 2 CC7.1 · SOC 2 CC7.2 · SOC 2 CC8.1 · SOC 2 A1.1
Operations Security Procedure
SOC 2 CC7.1 · SOC 2 CC7.2 · SOC 2 A1.1 · SOC 2 A1.2
Network Security Procedure
SOC 2 CC6.1 · SOC 2 CC6.6 · SOC 2 CC7.2 · ISO 27001 A.8.16
Communications and Network Security Policy
SOC 2 CC2.3 · SOC 2 CC6.6 · SOC 2 CC6.7 · ISO 27001 A.5.14
Encryption Policy
SOC 2 CC6.1 · SOC 2 CC6.7 · SOC 2 C1.1 · ISO 27001 A.5.17
Vulnerability Management and Patching Policy
SOC 2 CC4.1 · SOC 2 CC7.1 · SOC 2 CC7.2 · SOC 2 CC8.1
SDLC Procedure
SOC 2 CC3.4 · SOC 2 CC7.1 · SOC 2 CC8.1 · ISO 27001 A.8.4
System Acquisition and Development Lifecycle Policy
SOC 2 CC3.4 · SOC 2 CC8.1 · SOC 2 CC9.2 · ISO 27001 A.5.8
Change Management Policy
SOC 2 CC3.4 · SOC 2 CC7.1 · SOC 2 CC8.1 · ISO 27001 cl. 6.3
Data Protection Policy
SOC 2 P1.1 · SOC 2 P4.1 · SOC 2 P5.1 · SOC 2 P6.1
Privacy by Design Policy
SOC 2 P3.1 · SOC 2 P4.1 · SOC 2 C1.1 · SOC 2 CC8.1
Data Retention Policy
SOC 2 C1.2 · SOC 2 P4.2 · SOC 2 P4.3 · SOC 2 CC6.5
Vendor Management Policy
SOC 2 CC9.2 · SOC 2 CC3.2 · SOC 2 P6.1 · SOC 2 P6.4
Vendor Management Procedure
SOC 2 CC9.2 · ISO 27001 A.5.19 · ISO 27001 A.5.20 · ISO 27001 A.5.21
Incident Management Policy
SOC 2 CC2.3 · SOC 2 CC7.3 · SOC 2 CC7.4 · SOC 2 CC7.5
Incident Management Procedure
SOC 2 CC7.3 · SOC 2 CC7.4 · SOC 2 CC7.5 · ISO 27001 A.5.24
Data Breach Notification Policy
SOC 2 CC2.3 · SOC 2 CC7.4 · SOC 2 P6.3 · SOC 2 P6.5
Personal Data Breach Notification Procedure
SOC 2 CC7.4 · SOC 2 CC2.3 · SOC 2 P6.3 · SOC 2 P6.5
Business Continuity and Disaster Recovery Policy
SOC 2 A1.1 · SOC 2 A1.2 · SOC 2 A1.3 · SOC 2 CC7.5
Business Continuity Plan
SOC 2 A1.2 · SOC 2 A1.3 · SOC 2 CC7.5 · SOC 2 CC9.1
Subprocessors
Who handles customer data, and where.
A location is stated only where our own configuration pins it. Where we have not confirmed it with the vendor, it says "On request" rather than a guess. We give customers 30 days' notice before adding or replacing a subprocessor.
Hosting and storage 7
| Vendor | Purpose | Data processed | Location |
|---|---|---|---|
| Amazon Web Services | Hosting for the call platform: compute, load balancing, an execution store, and stored audio and documents. | What the platform processes while it runs (call details, phone numbers, transcripts, agent settings), plus stored audio, invoices and business-verification documents. | India (Mumbai) |
| Amazon Web Services: Secrets Manager | Stores API keys, and the carrier and push credentials that customers connect. | API keys; carrier account credentials; push-notification credentials. | United States (N. Virginia) for API keys; India (Mumbai) for carrier and push credentials |
| MongoDB Atlas | Primary database. | Agent settings; call records, transcripts and their analysis; contacts and campaigns, including phone numbers; the audit trail. | On request |
| Wasabi | Object storage for call recordings and voice-clone samples. | Call recordings; voice samples. | United Kingdom |
| Vercel | Hosts the website and dashboard and the endpoints the in-app SDK and web widget use; page-view analytics and speed measurements, with your consent. | Dashboard sessions; workspace users' account details; in-app session requests; page views. | India (Mumbai) for the dashboard's main API functions; others on request |
| Neon | The dashboard's database. | Workspace users and organisations; billing and invoices; workflows; consent records; the dashboard activity log; requests sent through our website. | On request |
| LanceDB Cloud | Search index for agent knowledge bases. | Text chunks and embeddings of the documents you upload. | United States (N. Virginia) |
AI, speech and voice 11
| Vendor | Purpose | Data processed | Location |
|---|---|---|---|
| Microsoft Azure OpenAI | Language model for the agents that use it, and for post-call summaries, translation, analysis and checks. | Conversation text and transcripts; prompts, including the variables you supply. | On request |
| OpenAI | Language-model fallback, and the dashboard's call evaluations, assistant and voice-studio helper. | Transcripts; prompts; evaluation conversations. | On request |
| Groq | Language model for agents that select it, and a last-resort fallback. | Transcripts; prompts. | On request |
| Sarvam AI | Speech-to-text, text-to-speech, language detection, transliteration, a language model when selected, and reading uploaded knowledge-base documents. | Call audio; transcripts; text to be spoken; uploaded documents. | On request |
| Deepgram | Speech-to-text (the platform's fallback), and text-to-speech when selected. | Call audio; text to be spoken. Deepgram's model-improvement program applies unless it is opted out of, and we have not opted out of it yet; the setting that opts out is built and being rolled out. | On request |
| Gnani | Speech-to-text when selected for an agent. | Call audio. | On request |
| ElevenLabs | Text-to-speech, and speech-to-text when selected. | Text to be spoken; call audio when used for speech-to-text. | On request |
| Cartesia | Text-to-speech, and voice cloning for the Voice Library. | Text to be spoken; recordings submitted for cloning. | On request |
| TimePay | Text-to-speech when selected for an agent. | Text to be spoken. | On request |
| Amazon Web Services: Polly | Fallback text-to-speech when an agent has no voice set. | Text to be spoken. | India (Mumbai) |
| mem0 | Agent memory for returning users, only when you turn memory on for an agent. | Redacted facts from conversations, keyed by a one-way hashed user id. | On request |
Telephony and call media 4
| Vendor | Purpose | Data processed | Location |
|---|---|---|---|
| jambonz.cloud | Call control and media for workspaces set up on the hosted voice network. | Call audio; phone numbers; call events. | On request |
| Exotel | Telephone carrier, on your Exotel account or ours. | Call audio; phone numbers; the carrier's own call recordings. | On request |
| Twilio | Telephone carrier, where one is set up for a workspace. | Call audio; phone numbers. | On request |
| Metered | Relays web and in-app calls when a direct connection isn't possible. | Encrypted call media in transit; IP addresses. | On request |
Accounts, billing and support 9
| Vendor | Purpose | Data processed | Location |
|---|---|---|---|
| Stytch | Sign-in for workspace users. | Names, email addresses, workspace membership and sessions. | On request |
| Brevo | Email: invitations, sign-in codes, approval notices, invoices, alerts and the replies to our website's forms. | Names; email addresses; invoice details. | On request |
| OneSignal | Browser notifications for the approvals inbox, for people who turn them on. | Your member id, your browser's notification subscription, and the notification's text, which can name a colleague and an agent or campaign. | On request |
| PayU | Payments for wallet top-ups. | Payer name, email address and phone number; the amount. | On request |
| Surepass | Checks a GSTIN during business verification. | GSTIN and registered business details. | On request |
| Crisp | Support chat on the website and in the dashboard, once you agree to it. | Your name, email address and chat messages. | On request |
| Google Analytics | Website analytics, when switched on and once you agree to it. | Page views; IP address; device details. | On request |
| ipapi.co | Shows the viewer's approximate location in Call History. | The viewer's IP address. | On request |
| Google Fonts, jsDelivr, Cloudinary, Tencent EdgeOne | Fonts and static files for the website and dashboard. | The visitor's IP address. | On request |
Monitoring 1
| Vendor | Purpose | Data processed | Location |
|---|---|---|---|
| Better Stack | Uptime heartbeats, and log storage when enabled. | Service logs, which can contain phone numbers and fragments of transcripts until masked logging is fully rolled out. | On request |
Providers a customer chooses for an agent
These process data only when you select them for an agent, and most need your own account or key.
- Alternative language models
- Anthropic, Google Gemini, Together AI, OpenRouter, AWS Bedrock, Cohere, DeepInfra, Perplexity, DeepSeek, Fireworks AI
- Alternative speech-to-text and text-to-speech
- AssemblyAI, Gladia, Soniox, Kalpa, Maya, Navana, Gemini, Azure AI Speech, Smallest.ai, Rime
- Language-switch judge for multilingual agents
- TypeSafe
- Web search in workflows and agent tools
- Firecrawl, Exa, Parallel, Tavily
Services on your own accounts
Your own carrier or SIP trunk, WhatsApp Business account, push-notification credentials, APIs, MCP servers and workflow integrations (a CRM, a calendar, a ticketing tool) run on your accounts. We act on your instructions when we use them, and they are not our subprocessors.
Tools that never handle customer personal data (source hosting, container registries and DNS) are not listed.
To be told before the list changes, email us. The terms are summarised on the DPA page.
Report a vulnerability
Found something? Tell us.
If you believe you have found a security issue in Graine, write to us. Please do not open a public issue or post it before we have had a chance to fix it.
How to report
Email founders@graine.ai with the subject “Security report”. Our contact details are also in security.txt.
Email a reportWhat to include
- What you found and where (a URL, an endpoint or a screen).
- The steps to reproduce it, and what you saw.
- Why it matters: what a person could do with it.
- How to reach you.
What we do, and what we ask
- We acknowledge a report within 2 business days.
- We keep you informed until it is fixed.
- Please test only against accounts you own, and stop at proof: do not read, change or keep anyone else’s data.
- Please do not run denial-of-service tests or social-engineer our staff or customers.
Last updated 4 November 2026. See also our data processing terms, our privacy policy and the security overview.