Trust

Security & trust

How Prynt protects your data, who else touches it, how long we keep it — and what is not done yet.

Encryption

TLS in transit. Secrets (sealing keys, Stripe keys, API credentials) are AES-256-GCM encrypted at rest under a key-encryption key derived from the application secret.

Authentication

Password hashing (PBKDF2), TOTP MFA, SSO via OIDC (PKCE) and SAML, tenant-scoped RBAC, and full JWT session revocation.

Your data stays yours

We act as your data processor under our DPA. Visitor IDs are scoped to your account and never linked across customers. We never sell personal data.

Auditability

Every privileged action is written to an audit log you can export as CSV for your SIEM.

Privacy controls

The browser agent withholds identification when the visitor sends Global Privacy Control or Do-Not-Track, and we support data export and right-to-erasure.

Reputation network

Accounts share one-way salted hashes of what they reported as fraud, and Prynt counts — on its own servers — devices and IPs seen at several accounts. An account only gets a flag and a count, never another account's data, and any account can stop contributing from the console.

Sub-processors

These are the only third parties that process data for Prynt. The API, the database and the console run on servers we operate ourselves in the United States, not on a third-party cloud. We will update this list before adding a new sub-processor.

ProviderLocationWhat forWhat it can see
Cloudflare, Inc.United StatesDNS, CDN, TLS termination and DDoS protection for pryntid.com and api.pryntid.com; Turnstile bot check on the sign-up, sign-in and partner formsRequest metadata in transit, including IP addresses
Stripe, Inc.United StatesSubscription billing and paymentsYour billing contact and payment details (not visitor data)
Twilio Inc. (SendGrid)United StatesTransactional email (account, security, billing, quota and digest emails)Recipient email addresses and email content
Anthropic, PBCUnited StatesAI features in the console: PAI (Replay session summaries) and Revenue AI draftsOnly the structured facts a feature needs (for PAI: a recording's pages, errors, clicks and frustration signals; for Revenue AI: company and activity data). Never raw recordings or typed input, and not used to train models
DeepSeekChinaThe chat assistant on pryntid.com and in the consoleOnly what a person types into the assistant, plus your plan and monthly usage when you use it in the console

Retention

Identification events, signal observations and visitors idle beyond your plan's window are deleted automatically every day. Your account, API keys, team and audit log are kept while your account is open.

PlanIdentification data kept for
Free30 days
Pro90 days
Scale365 days
EnterpriseAs agreed in your contract

Console sign-ins. We identify sign-ins to app.pryntid.com with Prynt itself (device ID, IP address, approximate location and network) to protect accounts and stop multi-account abuse. Each device record is deleted 24 months after the last sign-in from it, or when the person leaves the account. Details in the privacy policy.

IP addresses are stored in full in the cloud service today, because VPN, proxy and reputation signals need them. Dedicated deployments can store them truncated or not at all.

Backups of the database are taken daily, test-restored before they are kept, and held for 7 days on the API host and 14 days on a second server we operate. Data you delete leaves the backups when they expire.

The intelligence network

What is shared between accounts: one-way, salted hashes of the visitor IDs, IP addresses and account IDs (linkedId) that an account reported as fraud or abuse — through the outcomes API or «Mark as fraud» in the console — and nothing else. An entity an account only put on its deny list, or only blocked with one of its rules, counts for less and enters the network only when several accounts agree. Raw values, email addresses and identities are never exchanged, and a hash cannot be turned back into the value.

What Prynt counts across accounts: whether the same device (a hash of its on-device fingerprint), the same IP address or a newly appearing network (ASN) shows up at several accounts within 24 hours, and which bot user-agent tokens appear across them. This runs on our servers. An account only ever receives a flag and a number of other accounts — never which accounts, and never anything about their visitors. Visitor IDs stay scoped to the account that issued them and are not linked across customers.

Every account contributes by default and can stop at any time in the console (Rules & Risk › Reputation network). From the next rebuild, every six hours, its labels and its traffic are left out of both the shared hashes and the counts. Checking the network keeps working either way.

How the network works · Data Processing Agreement

Compliance status

Done: encryption at rest for secrets, MFA, SSO/SAML, RBAC, session revocation, audit-log export, automatic retention, right-to-erasure, data export, and our DPA.

Not done yet: Prynt has not completed a SOC 2 audit and has not had a third-party penetration test. When an auditor is engaged we will name them here with the target date.

Report a vulnerability

Please report security issues to [email protected]. See our security.txt. We respond within 72 hours.