Trust & Security

Your customer data stays yours.

How Kazozo handles the conversations, contacts, and content flowing through your agents — including what we store, what we don't, and what you can export on demand.

The three things we get right

Anything else is a detail. These are the three commitments we design every feature around.

Data minimization

We store what's needed to run your agents and nothing more. Every message is tied to a tenant, every action is logged, and you can export or delete any conversation through the dashboard.

Per-tenant isolation

Your customer data never crosses tenant boundaries. Agents are loaded with your policies, your FAQ content, and your routing rules — other businesses on the platform can't see any of it.

Model outputs are gated

AI responses flow through policy checks before they reach your customers. The model can't send SMS, push leads to your CRM, or take any action that isn't explicitly configured for your tenant.

How it works, in detail

Plain-English answers to the questions our customers ask before go-live.

Encryption

In transit: All traffic between your website, the Kazozo platform, and any integrations uses TLS 1.2 or higher.

At rest: Conversations, lead data, and configuration are encrypted at rest using industry-standard AES-256.

Access control

Every dashboard user is scoped to a single tenant. Internal access to production systems is restricted, logged, and audited. We don't read your customer conversations unless you open a support ticket that requires it — and we log when we do.

Retention & deletion

You control retention at the tenant level. Conversations can be set to expire after a defined window or kept indefinitely for training and audit. You can delete any conversation, lead, or contact on demand — including after a customer exercises a right-to-be-forgotten request.

No training on your data

Your customer conversations are not used to train the underlying AI models. They're used to run your agents — and only your agents. If we improve the platform based on patterns we see, those improvements are de-identified aggregate signals, never raw content.

Audit logs

Every agent action is logged with a timestamp, the inputs it saw, the decision it made, and the action it took. This is how we catch weird behavior before it reaches a customer — and how you'd reconstruct what happened if a customer complaint needed investigation.

Uptime & incidents

We target 99.9% platform uptime. When we miss it, we publish the incident — what happened, what was affected, and what we changed so it doesn't happen again. No corporate hedging.

What we aren't yet

Some SMB software vendors claim certifications they don't have. We'd rather be honest about where we are and where we're headed.

SOC 2 — in progress

We're building toward SOC 2 Type II and can share our current controls, subprocessor list, and DPA under NDA for customers doing formal vendor reviews. If that's a dealbreaker today, please tell us so we can prioritize it correctly.

HIPAA — not a BAA signatory

Kazozo is not HIPAA-covered today. For dental and medical customers, we help you configure agents so they never collect or store protected health information — appointment booking + intake happens through your existing HIPAA-covered systems, and Kazozo handles only public-facing conversations. If you have a HIPAA requirement that needs us to sign a BAA, we'd want to talk before you sign anything.

Questions we hear a lot

Who owns the customer data?
You do. The conversations, lead records, and contact info are yours. If you leave Kazozo, we help you export it and we delete our copy on request.
What happens to my data if I cancel?
Your account is retained for 30 days by default so you can come back or export anything you need. After that, conversation content is deleted. Billing records are kept for the period required by tax and compliance law.
Do you sell data to third parties?
No. Not for advertising, not for training, not for aggregated "insights" products. Your conversations exist to run your agents.
Where is data hosted?
In the United States by default. We can discuss region-specific hosting for enterprise customers with a data-residency requirement.
Can I request a DPA or subprocessor list?
Yes. Email hello@kazozo.com and we'll send you our current DPA, subprocessor list, and security overview under a mutual NDA.
How do you handle prompt injection?
Every model response runs through a policy layer before it reaches the customer. The model can't be tricked into performing actions outside its allow-list — no sending unauthorized SMS, no accessing other tenants, no leaking system prompts. The policy layer is versioned, tested, and audited separately from the model itself.

Still have security questions?

We'd rather walk through them with you on a call than leave you parsing a policy doc. No legalese — just real answers.

Book a demo → Email security@kazozo