Security and Control for AI Support.

Yes. Every organization's data — knowledge, conversations, leads — is isolated to its own tenant with row-level security. Your content is scoped to your account, not shared across the platform, and it trains nothing outside your own deployment.

No, and we won't claim otherwise. A formal SOC 2 attestation and HIPAA alignment are roadmap items, not current certifications. We draw a clear line between what's live today, what we scope per enterprise engagement, and what's on the roadmap.

Scraped and uploaded content is sanitized and the system prompt is hardened, so instructions hidden in a page or message can't hijack the bot. The crawler has SSRF defenses so it can't reach internal services, and the app has CSRF protection and layered rate limiting.

Yes. Sensitive actions are recorded in an audit log — who did what, and when. Combined with recorded conversations and their sources, you can trace both team actions and what the chatbot answered and where it came from.

Webhook URLs and secrets — like a Slack notification webhook — are encrypted at rest and masked when read back, so a configured integration doesn't expose its credentials in the interface.

Every organization's data — knowledge, conversations, leads, and customer interactions — is scoped to your tenant and not shared across the platform, enforced with row-level security at the database. The isolation is architectural, not a setting, and it's what makes multi-organization safe: several businesses under one login, each fully separated, with your content training nothing outside your own account.

Access is governed by roles, so people see and do only what their role permits. Owner, admin, agent, and viewer roles get appropriate scopes, and owners can define custom roles on a permission matrix — who manages knowledge, who works the team inbox, who views analytics, and who administers the account. See Team and Roles. Team access today is by role-based invitation.

Zurvo hardens the AI-specific and web attack surface, not just the perimeter. Scraped and uploaded content is sanitized before it reaches the model and the system prompt is hardened (prompt-injection hardening); the website crawler is defended against server-side request forgery so it can't reach internal services (SSRF protection); state-changing requests carry CSRF protection; and layered rate limiting spans chat, crawling, ingestion, and lead capture.

Customer data is encrypted in transit and at rest within the platform. Sensitive configuration like webhook URLs and secrets is encrypted at rest and masked on read. Zurvo also tracks AI usage and cost, giving you visibility into consumption rather than an opaque bill, surfaced in your analytics and dashboard usage meters.

These are not standard, self-serve, or certified features today. Single sign-on and SAML-based identity integration can be scoped as part of a managed enterprise engagement, and specific regional data-residency arrangements can be discussed for enterprise deployments rather than assumed out of the box. We'll be straight about scope and timeline and won't represent them as existing certifications; team access today is by role-based invitation.

A formal SOC 2 attestation and HIPAA alignment with Business Associate Agreements are roadmap considerations, not current certifications. We publish them because security teams find the gaps regardless, and a vendor who disclosed them upfront earns more trust than one who got caught. An accurate posture survives the review — an inflated one doesn't.