Security is not a feature we added — it is how TensorSound is built. Every architectural decision, from credential storage to breach response, is made with your data and your customers' data in mind.
TensorSound's core processing infrastructure — voice AI, telephony orchestration, database, and all call recordings — runs on servers physically located in the United Kingdom. Voice data, transcripts, and customer records are processed and stored in the UK and do not leave our infrastructure.
Our website is delivered globally via Cloudflare's CDN, which provides DDoS protection, edge caching, and DNS security as a separate layer in front of our origin servers. Sub-processors used for ancillary functions (transactional email, payment processing) are listed in Section 6 and documented in full in our MSA.
The following controls are implemented in production and applied uniformly across all customer tenants.
All credentials, API keys, and database passwords are held in HashiCorp Vault — a dedicated secrets management system used by banks and large enterprises. Nothing sensitive is stored in code, configuration files, or on disk.
User passwords are hashed using bcrypt at cost factor 12 before storage. The original password is never retained. Even in the event of a database breach, passwords cannot be recovered or reversed.
Customer data is isolated at the database layer using row-level security, enforced fail-closed — not just filtered in application code. A query that omits its tenant scope is rejected by the database itself, not silently returned.
Every backend service connects to the database with its own dedicated, minimally-scoped account — no shared credentials, no superuser access anywhere in the application layer. A compromise of one service cannot be used to reach another service's data.
Platform users access only the data their assigned role permits. Roles are enforced server-side on every request — not just in the interface. Privilege escalation attempts are blocked and logged automatically.
Authentication endpoints enforce progressive lockout. After repeated failed attempts, access is blocked for escalating durations — from one minute up to 24 hours — making automated credential attacks ineffective.
All connections use TLS 1.3. HTTP is automatically redirected to HTTPS. HTTP Strict Transport Security (HSTS) is enforced, preventing downgrade attacks. Connections not meeting this standard are rejected.
Every response includes Content-Security-Policy, X-Frame-Options, X-Content-Type-Options, Referrer-Policy, and Permissions-Policy — preventing clickjacking, MIME sniffing, and cross-site injection at the browser level.
Every payment event from Stripe carries a cryptographic signature we verify before processing. We do not store card numbers. Payment data is handled within Stripe's PCI-DSS certified environment.
TensorSound maintains a permanent, immutable audit trail across all platform activity. Every login, admin action, role change, and data access event is written to a tamper-evident log with a precise timestamp, user identity, IP address, and outcome.
In the event of an incident — whether internal or external — we can reconstruct exactly who accessed what, when, and from where. Logs are retained for a rolling 12-month period and are available to enterprise customers as part of their contractual audit rights.
Our authentication monitor detects anomalous patterns in real time — including repeated failed logins, access from unexpected locations, and privilege escalation attempts — and raises alerts for investigation.
Domixir Ltd (trading as TensorSound) is registered with the UK Information Commissioner's Office under registration number ZC108035. We are the data controller for data processed in connection with our platform and marketing website.
Compliance is built into the product, not applied retrospectively:
Our full privacy practices are set out in our Privacy Policy.
TensorSound is reviewed against two industry standards, not one — the general web application standard, and the standard specific to LLM-integrated systems. Most voice AI vendors are only evaluated against the first.
Full internal review across the complete platform — Go API, Rust context service, and Next.js frontend — against the current 2025 edition, including its newest categories (Software Supply Chain Failures, Mishandling of Exceptional Conditions). All identified findings remediated. Reviewed again as the standard updates, not a one-time exercise.
TensorSound's voice agent and RAG pipeline are separately reviewed against the standard built specifically for LLM-integrated systems — prompt injection resistance, tool-call authorisation, tenant isolation in retrieval, and output handling. A category most vendors in this space are never assessed against at all.
We use a small number of carefully selected third-party processors. Each is bound by a data processing agreement, and each transfer outside the UK is covered by an appropriate legal mechanism (UK IDTA, UK-US Data Bridge, or Standard Contractual Clauses).
| Processor | Purpose | Location | Transfer mechanism |
|---|---|---|---|
| Telnyx LLC | Telephony — voice calls and SMS | USA | UK-US Data Bridge / UK IDTA |
| Stripe | Payment processing | USA | EU-US Data Privacy Framework & SCCs |
| Cloudflare | CDN, DDoS protection, DNS | Global | Cloudflare UK DPA |
| SMTP2GO | Transactional email | Australia / USA | Standard Contractual Clauses |
| Netlify | Marketing website hosting | USA | Standard Contractual Clauses |
Core platform processing — AI orchestration, database, call recordings, voice infrastructure — runs exclusively on UK-based servers operated directly by Domixir Ltd. No voice data or customer records are transferred to any sub-processor.
Every TensorSound customer is covered by two documents that set out our security and data protection commitments in legally binding terms:
Enterprise customers may exercise their audit rights under the DPA. We will provide relevant documentation, configuration summaries, and — where applicable — third-party assessment reports in response to reasonable audit requests.
We take vulnerability reports seriously. If you believe you have found a security issue in TensorSound — on our website, platform, or API — we ask that you disclose it to us responsibly before publishing or sharing it publicly.
Email security@tensorsound.com with a clear description of the issue, steps to reproduce it, and any supporting evidence (screenshots, request logs, proof-of-concept). We will acknowledge your report within 2 business days and aim to resolve confirmed vulnerabilities within 30 days of validation, depending on severity.
We ask that you do not access, modify, or exfiltrate customer data; do not disrupt the service; and do not disclose the issue publicly until we have had a reasonable opportunity to remediate it. We commit to acting in good faith toward researchers who follow these guidelines.
Every control below is live in production today, not aspirational.
| Item | Status |
|---|---|
| UK GDPR compliance & ICO registration | Complete |
| PECR-compliant outbound calling controls | Complete |
| OWASP Top 10:2025 & OWASP LLM Top 10:2026 review & remediation | Complete |
| Security headers (CSP, HSTS, X-Frame-Options, COEP) | Complete |
| Secrets vault — HashiCorp Vault, per-service least-privilege access | Complete |
| Immutable audit logging | Complete |
| Database-layer tenant isolation (row-level security) | Complete |
| Internal adversarial testing — session/tenant boundary tampering | Complete |
| Automated CI scanning — every change, before it ships | Complete |
| Continuous automated security scanning | Ongoing |
We can share detailed technical security documentation under NDA — architecture diagrams, control implementation detail, and our current compliance roadmap — so your team can complete its review without waiting on a questionnaire back-and-forth.
For security enquiries, vulnerability reports, DPA requests, or compliance questions: