Security and data protection at TruSend
This page describes the technical and organizational measures TruSend applies to personal data, payments and audit records, both in the chat widget on your customers’ site and in the workspace your team uses.
This page is provided for information. The security measures we commit to are those in Annex II of our Data Processing Agreement; this page does not form part of the Data Processing Agreement or the Terms of Service.
Last reviewed · October 16, 2026
How a customer message is processed
The main stages a customer message passes through, and the control applied at each. The providers that process data are listed on the subprocessors page.
- 1
No visitor sign-in required
The widget opens a session without requiring the visitor to sign in. Bootstrap requests are sent without credentials, so cookies on your site are not attached to them. To continue the conversation, the widget sets a first-party visitor identifier in the visitor’s browser; this identifier is personal data.
- 2
Encrypted in transit
The widget connects over a secure WebSocket, and its bootstrap requests use HTTPS. Public connections and connections to the database use TLS; traffic between our servers runs through WireGuard tunnels.
- 3
PII redaction before the model
Email addresses, phone numbers, card numbers and IP addresses detected in content sent to AI models are removed before it is sent. This includes content sent for replies, translation, conversation summaries and knowledge-base indexing and search. Contact details that your workspace publishes for customers in its knowledge base are not removed. Images, voice messages, videos and scanned documents that customers send are sent to an AI model as they are; the removal applies to the text produced from them.
- 4
Scoped to your workspace
Stored encrypted at rest. In the application, queries are filtered by the organization of the authenticated connection.
- 5
AI reply generation
AI replies are generated through OpenRouter, our subprocessor in the United States, under its data processing agreement. Our OpenRouter account is configured to route requests only to the upstream providers listed on our subprocessors page, which are its subprocessors. We do not use customer data to train AI models, and our OpenRouter account allows only providers that do not train on submitted content.
- 6
Written to the audit trail
Settings, account and AI configuration changes are logged with who made them, when, and the values before and after, and are retained for 90 days; changes to contact retention policies and legal holds are retained for 7 years.
The SDK on your site and the desk your team works in
TruSend has two surfaces: the chat widget your customers use and the workspace your team uses. The controls for each are listed below.
Nerv
The widget on your siteThe embed script contains public identifiers only, not secret keys
- No signing key in your page: the widget bundle contains public identifiers only, and the build excludes signing secrets.
- Domain allowlist: when enabled, an origin outside the list fails verification, and the widget does not load your configuration or open a connection.
- Content sanitized before rendering: message markdown and help-center content pass through allowlist-based sanitizers, and javascript:, data: and file: links are blocked.
- Renders in a shadow root, not in an iframe. The shadow root separates the widget’s styles from your page’s stylesheets; it isolates styling only and is not a security boundary against scripts running on your page.
- Crash reports are scrubbed in the visitor's browser before they are sent to our error-monitoring provider: password, token and cookie fields, and values matching an email-address pattern, are removed.
- Optional signed identity: the widget accepts a token signed by your backend to link a conversation to a known customer.
Portal
Where your team worksRole-based access, and an audit log of settings and account changes
- Tenant isolation in the application: your organization is taken from the authenticated connection, not from the request, and queries are filtered by it.
- Role-scoped access, with locked workspace defaults enforced by the server: overrides of a locked field are rejected.
- Settings, account and AI configuration changes are logged with who made them, when, and the values before and after, and are retained for 90 days; changes to contact retention policies and legal holds are retained for 7 years.
- API keys are displayed masked; each workspace can hold two active keys at the same time.
- PII redaction: email addresses, phone numbers, card numbers and IP addresses detected in content sent to AI models are removed before it is sent. This includes content sent for replies, translation, conversation summaries and knowledge-base indexing and search. Contact details that your workspace publishes for customers in its knowledge base are not removed. Images, voice messages, videos and scanned documents that customers send are sent to an AI model as they are; the removal applies to the text produced from them.
- Encrypted at rest; unsent drafts are encrypted in the browser with an AES-256-GCM key issued by our server, held only in memory and valid for 8 hours.
GDPR and PCI-DSS
The summaries below describe how our controls relate to these frameworks. They are not a certification: TruSend does not hold a certification under either.
EU General Data Protection Regulation
We act as processor for the conversations, contacts and knowledge-base content in your workspace, and as controller for your account, workspace configuration and billing data. Data-subject rights are supported through self-service features in the product.
- Self-service data export: any user can download their personal data as a portable archive.
- Right to erasure with a two-step email confirmation: the sign-in account and profile data are deleted, and the user record that remains linked to past messages no longer contains the name or email address.
- Records of processing activities are kept for both roles, controller and processor, and are made available to a supervisory authority on request.
- PII redaction: email addresses, phone numbers, card numbers and IP addresses detected in content sent to AI models are removed before it is sent. This includes content sent for replies, translation, conversation summaries and knowledge-base indexing and search. Contact details that your workspace publishes for customers in its knowledge base are not removed. Images, voice messages, videos and scanned documents that customers send are sent to an AI model as they are; the removal applies to the text produced from them.
- The Data Processing Agreement and the subprocessor list are published at trusend.ai/dpa and trusend.ai/subprocessors.
Payment Card Industry Data Security Standard
TruSend does not process card payments itself. Subscription payments are processed by Paddle as merchant of record.
- Subscription payments are processed by Paddle, our merchant of record and a PCI-DSS Level 1 service provider. TruSend does not receive or store the card details used to pay for a subscription.
- Settings, account and AI configuration changes are logged with who made them, when, and the values before and after, and are retained for 90 days; changes to contact retention policies and legal holds are retained for 7 years.
- A detector in the workspace reply composer blocks replies in which it detects card numbers, IBANs, US Social Security numbers or Russian SNILS numbers, and the server repeats the check. Internal notes are excluded.
- Card numbers that your customers type into a conversation are stored as message content.
- API keys are displayed masked and can be rotated or revoked.
Platform security controls
These measures are included in all plans.
Encrypted in transit and at rest
Public connections use TLS, and traffic between our servers runs through WireGuard tunnels. Stored data is encrypted at rest. Unsent drafts are encrypted in the browser with an AES-256-GCM key issued by our server, held only in memory and valid for 8 hours.
Isolated by tenant
In the application, your organization is taken from the authenticated connection, not from the request, and queries are filtered by it.
PII redaction before the model
Email addresses, phone numbers, card numbers and IP addresses detected in content sent to AI models are removed before it is sent. This includes content sent for replies, translation, conversation summaries and knowledge-base indexing and search. Contact details that your workspace publishes for customers in its knowledge base are not removed. Images, voice messages, videos and scanned documents that customers send are sent to an AI model as they are; the removal applies to the text produced from them.
Sensitive numbers blocked on send
A detector in the workspace reply composer blocks replies in which it detects card numbers, IBANs, US Social Security numbers or Russian SNILS numbers; internal notes are excluded. The server repeats the check, and the audit event records the type of data detected, not the value.
Audit trail for settings and account changes
Settings, account and AI configuration changes are logged with who made them, when, and the values before and after, and are retained for 90 days; changes to contact retention policies and legal holds are retained for 7 years.
Data export and deletion
Contacts and companies can be exported from the workspace as CSV files, and each user can export their own personal data. When the service ends or you switch providers, other data is returned as the Data Processing Agreement and the Terms provide. A deleted workspace and the files stored with it are deleted from production systems 30 days after the workspace is deleted, a period during which it can be restored, and from database recovery history within a further 7 days. We do not use customer data to train AI models.
Certifications and commitments
- Describe the controls the product ships.
- Describe how our features relate to GDPR and PCI-DSS requirements.
- Publish the Data Processing Agreement and the subprocessor list.
- Claim a SOC 2 report or ISO 27001 certification. TruSend does not currently hold either.
- Use customer data to train AI models.
Common questions
Certifications, hosting, the Data Processing Agreement and access control.
No. TruSend does not currently hold a SOC 2 report or ISO 27001 certification. Our security measures are described on this page and in Annex II of the Data Processing Agreement.
Bring this to your security review
Our Data Processing Agreement and subprocessor list are published. For questions about our controls, email us.
Found a vulnerability? Report it to security@trusend.ai.