Security
Serious data protection on every plan, from $49 a month.
Encrypted data, kept separate from every other customer, with two-step sign-in and built-in privacy tools. No security add-on to buy.
Built in on every plan
Encrypted data
Login details, emails and conversations are encrypted when stored and when sent.
Your data stays yours
The database itself blocks any customer from seeing another's records, and automated tests check it.
Two-step sign-in
Admins can use an authenticator app, and you can require it for your whole team.
Privacy tools
Export or erase a visitor's data yourself, with a record of every request.
Watches for misuse
Every chat is checked for attempts to trick the AI, and we monitor for spikes.
Signed lead deliveries
Leads sent to your other apps are signed, so fake leads can't slip in.
See how the privacy tools work in the lead inbox and how signed deliveries reach your apps.
For technical buyers: how it works
- Encryption: AES-256-GCM at rest with versioned ciphertext for zero-downtime key rotation. TLS for every connection.
- Tenant isolation: Postgres row-level security with deny-by-default policies on every tenant table. Every query is scoped by tenant, and automated tests attempt cross-tenant queries to confirm they fail.
- MFA: TOTP through standard authenticator apps. Existing sessions are invalidated when MFA is turned on.
- GDPR: self-serve export (signed JSON download) and erasure by email or session ID, with audit log entries.
- SSRF protection: the website reader and webhooks refuse private and internal addresses, including localhost, RFC 1918 ranges and cloud metadata endpoints.
- Prompt-injection monitoring: visitor messages are scanned for 8 injection patterns and replies for 5 kinds of leakage. Detections are logged without blocking the chat, and high-severity patterns alert our team.
- Webhooks: HMAC-SHA256 signatures in the
X-Wengrow-Signatureheader. - Edge: Cloudflare with firewall rules, bot management on API paths, and security headers such as HSTS and CSP.
What we don't have yet
We'd rather tell you upfront.
No SOC 2 yet
We're working toward a SOC 2 audit. Our controls are documented but not yet audited. If you need SOC 2 today, NightMaven isn't the right fit yet.
Not for health records
NightMaven isn't HIPAA-certified. It's fine for marketing and booking chats before a patient shares health details, but not for patient health information.
No ISO 27001 yet
Documented, not audited. It comes after SOC 2.
Running a clinic? See what NightMaven can and can't do for healthcare.
For teams with stricter rules
Single sign-on
Pro and up. Sign in through Okta, Microsoft Entra, Google Workspace or any SAML provider.
Your own rules
Business+. Write instructions the agent must always follow.
Careful personality
Every plan. A preset tuned for finance, healthcare and legal.
Full chat records
Every message is saved with times and sources, so your compliance team can review any conversation.
See the single sign-on setup, agent settings, and how financial advisors use NightMaven.
Security FAQ
- Not yet. We're working toward a SOC 2 audit. The controls, including encryption, access management, data separation, two-step sign-in and incident response, are in place and documented but not yet audited. If you need SOC 2 today, NightMaven isn't the right fit yet.
- No. NightMaven isn't HIPAA-certified and doesn't sign business associate agreements. It's fine for marketing, booking and screening chats before a patient shares health details.
- The main database is hosted in the US (US-East). The AI runs through Anthropic and Google. Search data is stored with Pinecone and email is sent through Resend. Every provider is listed in the privacy policy.
- Only you and your team. NightMaven engineers can access it only to respond to an incident, and every access is logged. We never use your data for anything else.
- No. Our agreements with Anthropic and Google don't allow training on your data, and we don't train our own models on it either.
- In your admin, go to Privacy, then Data Requests. Enter the visitor's email or chat session and confirm. It's erased right away and logged. Exports work the same way, with a secure download link.
- We follow a written incident plan. If customer data is affected, we tell affected customers within 72 hours, sooner for serious issues, with what happened and what we did.
- We run continuous automated security scanning. Third-party penetration testing is part of our SOC 2 work and isn't done yet. If a pentest report is a must-have, NightMaven isn't the right fit yet.
Enterprise-level protection at small-business prices.
Every security feature is on, on every plan. No add-on to buy.