Trust center
Written for the person whose job is to say no.
Security reviews run on specifics. Here are ours, including the parts still in progress.
This page covers the platform Taiga runs. How the software we deliver is governed lives on /governance.
There are no database passwords. None exist to steal.
The API authenticates to the database with short-lived IAM tokens. Taiga system description, section 7, available to your security team on request.
No SSH. No bastion. No standing administrative paths.
Administrative access is an on-demand task through AWS SSM, opened for the job and gone after it. Taiga system description, section 5, available to your security team on request.
The audit archive is immutable for seven years. We could not edit it if we wanted to.
CloudTrail, delivered to an isolated archive account under S3 Object Lock in COMPLIANCE mode. Taiga system description, section 11, available to your security team on request.
The specifics
What a security review finds.
Where your data lives
EU by policy, not by promise. Customer data at rest stays in AWS eu-central-1, Frankfurt. Model inference through Amazon Bedrock stays inside EU regions. The mechanism is an AWS Control Tower landing zone, organization-wide guardrails, SCPs preventing security-control tampering, a Resource Control Policy data perimeter, and region controls. Not a configuration a project team can forget or override.
The fine print, before you ask. Inference may run in more than one EU region for capacity and resilience; nothing persists there and nothing leaves the EU. The CDN's TLS certificate is managed in a US region because the provider requires it. Certificate metadata only, never customer data.
Compliance status
ISO 27001 in progress, SOC 2 planned. That is a status report: we hold neither certificate today, and you will not see those logos here until we do. We act as processor under a standard Data Processing Agreement. The platform runs on AWS, whose own ISO 27001, SOC 1, 2 and 3, and C5 reports are available through AWS Artifact.
Tenant isolation, three layers
Row-level security in PostgreSQL scopes every query to your tenant, enforced in the database, and the application role cannot bypass it. Uploaded documents live under per-tenant S3 prefixes. Retrieval filters by tenant at query time, so your documents never surface in another tenant's context.
Encryption
TLS 1.2+ in transit, everything encrypted at rest. Customer-uploaded documents and their embeddings get a dedicated customer-managed key with automatic rotation. Unencrypted or public storage is denied by policy, organization-wide.
Models and guardrails
Anthropic Claude via Amazon Bedrock, inference inside EU regions. Amazon Bedrock Guardrails run on agent traffic: sensitive PII is blocked or anonymized, and prompt-injection detection covers user-facing agents.
Detection and posture
Amazon GuardDuty watches every account. AWS Security Hub scores the environment against CIS and AWS foundational benchmarks. Amazon Inspector continuously scans containers and functions for known vulnerabilities. Every pull request to the platform's main branch runs static analysis, secret detection, and dependency scanning. All infrastructure is peer-reviewed code, so every change leaves a history.
Edge protection
AWS WAF fronts the API with managed rule sets for common attacks, SQL injection, and known bad inputs, plus per-IP rate limiting. CloudFront enforces HTTPS and security headers at the edge.
Identity
Amazon Cognito manages authentication, so Taiga never stores or sees a raw password. OAuth 2.0 authorization code flow only, short-lived tokens with revocation enabled, and adaptive risk detection in enforced mode.
Retention and deletion
Database backups are automated and point-in-time. Uploaded document versions expire on a defined schedule, and operational logs are retained on defined schedules by type. Account data lasts the length of the contract: deleted on termination or on request, under the Data Processing Agreement. The audit archive follows its own rule, immutable for seven years.
Availability and backups
The application and database tiers run across multiple availability zones. Backups are automated and encrypted, and copies land in an isolated backup account, so a compromise or an accidental deletion in a workload account cannot take the backups with it. Restore is point-in-time. Deletion protection is enabled on the production database, and teardown takes a final snapshot before anything is removed.
Subprocessors
Who touches your data.
Swipe for all columns
| Subprocessor | Purpose | Data exposure |
|---|---|---|
| Amazon Web Services | Hosting, database, storage, identity, AI inference, email | All platform data, eu-central-1 Frankfurt |
| Anthropic via Amazon Bedrock | LLM inference | Prompts and responses, inside EU regions, never used for training |
| GitHub | Source-code access through a customer-installed GitHub App | Customer repositories, scope controlled and revocable by the customer |
| Optional SSO identity provider; documentation lookups by the design and security agents | SSO identity if enabled; agent documentation queries |
system description · section 13 · request only
The full subprocessor catalogue ships as an appendix to the Data Processing Agreement.
Direct questions
Questions your security team will ask.
Which models does Taiga use?
Anthropic Claude via Amazon Bedrock, inference inside EU regions. The models are accessed exclusively through Bedrock, nothing self-hosted and no second provider.
Is our data used to train models?
No. Amazon Bedrock does not use prompts or outputs to train foundation models, and an organization-wide policy opts Taiga out of AWS AI services using its content. Zero customer data is used for model training, ours or anyone else's.
How is tenant data isolated?
At three layers. Row-level security scopes every database query to your tenant. Uploaded documents live under per-tenant S3 prefixes, encrypted with a dedicated customer-managed key. Retrieval filters by tenant, so your documents never appear in another tenant's context.
Where do our code and data live?
Your code stays in your GitHub behind an App you install and can revoke. The platform never holds a copy of your repo. Built applications deploy to your cloud account, or a Taiga-managed one. Customer data at rest stays in AWS Frankfurt.
Can Taiga run without internet access?
The platform is a managed EU SaaS, so no, and we will not pretend otherwise. The systems it delivers deploy into your environment under your network policy. If a managed platform is off the table for you, tell us what is on it.
The detailed system description exists for your security team. Ask and we send it.
Written for your security team. Fifteen sections, from tenant isolation to the 7-year immutable audit archive. Leave your work email. A person sends it, never an autoresponder.
Ask the question this page did not answer. hello@tai.ga
Our claims are versioned like our infrastructure. /receipts
Security findings go to the address in our security.txt. We respond, we credit, we fix.
Bring the person who will say no.
Book a demo