AI Your Customers Can Trust With Their Data
Tridan AI protects customer data with PII detection and anonymization before any request reaches a model provider, encryption in transit and at rest, isolated multi-tenant organizations with roles and permissions, and a choice of where the platform runs: managed cloud, your VPC or on-premises with local models via Ollama. A DPA is available for every customer.
How does Tridan AI protect customer data?
Tridan AI protects customer data in four ways: personal data is detected and anonymized before it leaves the platform, conversations and documents are isolated per organization with role-based access, traffic is encrypted in transit and stored encrypted at rest, and you choose the deployment, managed cloud or self-hosted, so data residency follows your policy rather than ours.
Read the Data Processing AgreementWhat protects personal data in conversations?
Four layers work together. PII detection and anonymization mask names, phone numbers and other identifiers before a prompt reaches an external model. Self-hosting keeps the whole platform, including local models, inside your network. Your content is never used for training. Multi-tenant isolation with roles and permissions limits who in your organization can see which conversations.
PII Detection and Anonymization
Personal data in conversations is detected and masked before any request reaches an external model provider.
Self-Hosted: VPC or On-Premises
Run the entire platform inside your own infrastructure, including local models via Ollama, so conversations never leave your network.
Never Used for Training
Your conversations, documents and knowledge bases stay yours. They are not used to train AI models.
Multi-Tenant Isolation
Isolated organizations with roles and permissions keep teams, brands and clients cleanly separated.
Which security practices are in place?
Security on Tridan AI is a set of practices you can verify in the contract and in the product: encryption in transit and at rest, PII anonymization, data residency options, a signed Data Processing Agreement, self-hosting, and access control through roles and permissions. Each one is described below and in the legal documents.
Encryption in transit and at rest
Traffic is encrypted with TLS; stored conversations, documents and credentials are encrypted at rest.
PII anonymization
Personal identifiers are detected and masked before prompts reach external model providers.
Data residency options
Managed cloud, your VPC or on-premises: you decide the network and country your data stays in.
Data Processing Agreement
A DPA covering roles, purposes, security measures and deletion is available to every customer.
Self-hosting
The full platform, including local models via Ollama, can run inside your own infrastructure.
Access control
Organizations are isolated; roles and permissions decide who sees conversations and the handoff inbox.
What should you ask any AI vendor about data?
Ask three things before deploying AI on customer conversations: where every prompt and transcript is stored and which third parties receive it, whether data can stay in your own region or network, and how a person takes over when the AI should not answer. Tridan AI answers all three on this page and in the DPA.
Your customers' data is in every prompt
Every support chat and phone call carries names, addresses and order details. An AI vendor must account for exactly where that data goes.
Data residency is not optional for everyone
Some industries cannot send customer conversations to a third-party cloud. Self-hosting keeps residency under your control, not your vendor's.
Humans must stay accountable
Every escalation lands in the human handoff inbox with a full transcript, so your team can always see and own what the AI said.
Building AI governance in your organisation? Tridan Academy, a separate training company, offers EU AI Act and ISO/IEC 42001 training.
Where can Tridan AI run?
Tridan AI runs in three places with the same features: the managed cloud (the Tridan AI platform dashboard), your own VPC, or fully on-premises. Managed cloud is the fastest start and includes PII protection. Self-hosted options keep every conversation, document and model call inside your perimeter, with local models via Ollama when no external provider is allowed.
Swipe sideways to see the full table
| Managed cloud | Self-hosted (VPC / on-premises) | |
|---|---|---|
| Where data lives | Tridan AI managed cloud, isolated per organization | Your VPC or your own servers; nothing leaves your network |
| Models | OpenAI, Anthropic, Meta and DeepSeek through provider APIs | The same providers, or local models via Ollama with no external calls |
| PII masking | Built in; identifiers are anonymized before prompts leave the platform | Built in; applies to cloud providers and local models alike |
| Who operates it | Tridan AI runs and monitors the platform | Your team, with Tridan AI installing and supporting it |
| Updates | Rolled out continuously by Tridan AI | Scheduled with your team, in your change windows |
| Best for | A fast start with no infrastructure to manage | Regulated industries and strict data residency requirements |
Start on the managed cloud; move to self-hosted when a data residency or compliance requirement demands it.
Security Questions, Answered Plainly
Where is my data stored?
On the managed cloud, your conversations, documents and knowledge bases are stored in your own isolated organization on infrastructure operated by Tridan AI, encrypted at rest. If you need a specific country or network, choose a self-hosted deployment: the entire platform then runs in your VPC or on your own servers, and nothing is stored outside your infrastructure. The Data Processing Agreement describes how and where data is processed under your contract.
Is my data used to train AI models?
No. Your conversations, uploaded documents and knowledge bases are used only to answer your own customers and team; Tridan AI does not use them to train AI models. When an agent calls an external provider such as OpenAI or Anthropic, PII anonymization can mask personal data first, and the request is sent through the provider's business API under that provider's terms. If you want no external calls at all, run local models via Ollama on a self-hosted deployment.
Can I run Tridan AI in my own VPC?
Yes. The full platform, including the Chat AI Agent, the Call Center Voice Agent, the handoff inbox and the integrations, can be deployed in your own VPC or on-premises. You can keep using cloud model providers through the anonymization layer, or run local models via Ollama so no request leaves your network. Tridan AI installs it, agrees an update schedule with your team and supports it; the self-hosted page explains the setup in detail.
How is personal data anonymized?
Before a prompt is sent to a model provider, the platform scans the text for personal data such as names, email addresses, phone numbers and similar identifiers, and replaces them with placeholders, so the provider never receives the real values. The detection applies to chat conversations and call transcripts alike, and it is one of the reasons the platform is used by teams that handle sensitive customer data. On a self-hosted deployment it also applies to local models.
Do you sign a DPA?
Yes. Tridan AI provides a Data Processing Agreement that sets out the roles of controller and processor, the purposes of processing, confidentiality, security measures, assistance with data subject requests and deletion at the end of the contract. It is published on the legal pages and is signed as part of your agreement; customers with specific regulatory requirements can discuss amendments with us before going live.
Who can access my workspace?
Only the people you invite. Each organization is an isolated tenant; within it, roles and permissions decide who can edit agents, upload knowledge, view conversations and work the handoff inbox. On the managed cloud, Tridan AI operates the infrastructure under the Data Processing Agreement and the privacy policy. On a self-hosted deployment, access is governed entirely by your own infrastructure and identity systems.
What happens when a conversation is escalated to a human?
The AI agent stops answering the sensitive part, tells the customer that a person will follow up, and places the full transcript in the human handoff inbox. Only staff with the right permissions see it there, and they reply from the same workspace, so the customer never has to repeat themselves. Escalation rules are set per agent, for example for billing, complaints or legal questions.
Last updated · Reviewed by Nemanja Milivojevic, Founder
Bring Your Security Questions
Put your infrastructure, privacy and data residency questions to us on a call. Bring your technical people; we will bring ours.