Trust & Security

VWAM is a hosted AI commerce suite. It stores and processes data for the merchants that use it. That data covers product catalog records, conversation records from the Shopping Assistant, and answer-engine visibility results from the AEO Tracker. This section explains the architecture that protects the data. It covers the storage regions, the content that VWAM sends to the AI providers, the access controls, and the limits on assistant responses.

This section is written for the people who evaluate VWAM for a merchant. The audience is a security, privacy, or platform team that performs an architectural review. These pages do not cover every implementation detail. For data-model specifics, sub-processor data processing agreements (DPAs), our current audit posture, or a geo-residency commitment for a specific contract, contact us.

VWAM is an independent platform. VWAM integrates with BlueConic and with other systems. A merchant opts into each integration. An integration is not part of VWAM's own storage or processing.

The short version

• Hosting. VWAM runs on AWS. Production runs in the US region us-east-1. A dedicated EU chat environment runs in the Frankfurt region eu-central-1. That environment does not serve shopper traffic yet, so VWAM handles EU conversations in us-east-1 today.
• Encryption. VWAM encrypts all data at rest and in transit. Databases use AWS KMS keys. Object storage uses server-side encryption. Customer-facing endpoints require TLS 1.2 or higher, and redirect plaintext HTTP to HTTPS.
• Isolation. VWAM partitions data per organization. Every record is scoped to the organization that owns it. The platform enforces that scope on every API request.
• Access control. The platform uses role-based access control (RBAC) with signed, expiring tokens. Internal service-to-service calls use a separate channel with key authentication.
• AI providers. OpenAI and Anthropic power the assistant as sub-processors, through their standard APIs. The AI providers never use conversations to train their models. Our primary reasoning path opts out of provider-side retention.
• Guardrails. The assistant answers from each merchant's own product catalog and knowledge base. A fixed set of tools and a built-in system prompt bound its responses. It recommends only real products, and it stays on topic.
• Privacy program. Automated data-subject request tooling for erasure and export is in place. We are finalizing retention schedules, EU geo-residency guarantees, and EU consent controls.

In this section

• Data storage & geo-residency — what VWAM stores, the storage region, and the difference between the US and EU deployments today.
• AI providers & how prompts are handled — the content that VWAM sends to OpenAI and Anthropic, and how their retention works.
• Access control & tenant isolation — authentication, RBAC, per-organization scope, and internal APIs.
• Response guardrails — the controls that keep the assistant grounded, on topic, and constrained.
• Privacy & compliance — data-subject requests, our GDPR program, and sub-processors.