Security Policy

Zrico is a private AI workspace for law firms, accounting firms and insurance brokers. This page explains how we protect your firm's data, and where our protections stop. We would rather be specific and honest than impressive.

Summary

• Every firm gets its own isolated workspace. Firms cannot reach each other.
• All traffic passes through Cloudflare. Our servers accept connections from Cloudflare only.
• Sign-in is invite-only. Two-factor authentication (2FA) is offered to every user.
• AI requests are processed in the United States only. No AI provider trains on your data.
• Our AI gateway does not store your prompts or responses.
• We do not hold SOC 2, ISO 27001, HIPAA or any other certification. Do not use Zrico for patient health information.

An isolated workspace for each firm

Each firm runs on its own stack, not on shared tables in one big database. Your firm gets:

• its own LibreChat application at https://{slug}.zrico.app
• its own MongoDB database for chats and settings
• its own search index (Meilisearch)
• its own vector database for document search
• its own Docker networks

The networks are set up so that one firm's workspace cannot connect to another firm's. Databases, search indexes and internal services are never exposed to the internet.

Workspaces live on zrico.app, a separate domain from our website and dashboard at zrico.com. The two never share cookies.

Network edge

• Cloudflare sits in front of every workspace and our dashboard. It provides TLS, a web application firewall (WAF) and DDoS protection.
• Origin lockdown. Our servers accept web traffic only from Cloudflare's published IP ranges. Direct connections to the server are refused.
• CrowdSec watches for abusive behavior (such as password guessing and scanning) and blocks the source.
• Administrative SSH access is limited to a short list of administrator IP addresses.

Access and sign-in

• Invite-only. At launch, staff sign in with an email invitation from their firm's owner or admin. There is no public sign-up.
• 2FA. Every workspace user can turn on two-factor authentication. We recommend that every firm require it of its staff.
• Dashboard. The account dashboard at app.zrico.com uses email and password with email verification. 2FA (an authenticator app) is required for firm owners and admins.
• Microsoft 365 sign-in (coming soon). Firms on Microsoft 365 will be able to sign in with Microsoft Entra ID, limited to their own tenant, the users they assign and their allowed email domains. Their Entra policies, such as multi-factor authentication and conditional access, then apply. This is not available at launch.
• Our staff. Only a small number of Zrico administrators can reach production systems, and they must use 2FA. We access a firm's workspace data only when needed to support that firm, run the service or meet a legal obligation.

AI processing

• Prompts and files you send to a model go through our AI gateway to the model provider you choose: Anthropic (Claude, with Amazon Bedrock as a fallback) or OpenAI (GPT).
• All AI processing happens in the United States.
• No provider trains models on data sent through our API accounts.
• Our gateway does not store prompts or responses. It records only usage data: the user's email, the model, token counts and cost. We keep those usage records for 400 days.
• Providers keep data for limited periods for abuse monitoring. The details, provider by provider, are on our AI data use page.

Encryption

• In transit. Connections between your browser and Cloudflare, and between Cloudflare and our servers, use TLS 1.2 or higher. Connections from our gateway to AI providers use TLS.
• Backups. Workspace data is backed up every night. Backups are encrypted before they leave the server (using restic) and stored in Cloudflare R2.
• Secrets. Database passwords and encryption keys for each workspace are stored only on the server, readable by root only. Copies kept off the server are encrypted (SOPS with age).

Data retention

• Chat retention is your firm's choice. Owners can keep chats indefinitely or delete them automatically after 30, 90 or 365 days.
• Backups age out. We keep 7 daily, 4 weekly and 12 monthly backups. So data you delete can remain inside encrypted backups for up to about 12 months before it is gone for good.
• Cancellation. When a firm's cancellation takes effect (at the end of the free trial or of the paid month), its workspace is suspended and deleted 30 days later. We offer an export by email during those 30 days.

Logging

We log what we need to run and secure the service: request times, status codes, errors, sign-in events and usage totals. Our logs never contain passwords, secrets, prompts, responses or file contents.

Patching and software supply chain

• Servers apply operating system security updates automatically (unattended upgrades).
• Every container image is pinned to an exact, reviewed version. We never run "latest" or development builds in production.
• We scan images for known vulnerabilities with Trivy and Grype, and we update promptly when a fix for a serious issue is available.
• Secrets are kept off source control and are encrypted when stored off the server.

Monitoring and status

We monitor uptime, server health and error rates, with alerts sent to our team. Live service status and incident updates are posted at https://status.zrico.com.

Incident response

We have a written plan for detecting, containing and recovering from security incidents, and for telling affected customers. If a breach affects your firm's data, we will notify you without undue delay and within 72 hours of becoming aware of it. Read our incident response policy.

Reporting a vulnerability

If you believe you have found a security issue in Zrico, email security@zrico.com. Please include enough detail for us to reproduce it. Our machine-readable contact file is at https://zrico.com/.well-known/security.txt.

We ask that you:

• give us reasonable time to fix the issue before you disclose it
• do not access, change or delete other people's data
• do not degrade the service for others (no denial-of-service testing)
• do not use social engineering or physical attacks

We will acknowledge your report, keep you updated, and credit you if you wish. We do not currently run a paid bug bounty.

What we do not have

We want you to know our limits before you rely on us.

• No certifications. We do not hold SOC 2, ISO 27001, HIPAA or any other security certification or attestation, and we have not had a third-party audit.
• Not for health information. We do not sign HIPAA business associate agreements. Do not put patient health information in Zrico.
• Not for regulated securities communications. We do not provide a compliant archive for broker-dealer or RIA communications.
• No zero data retention yet. AI providers keep data for limited periods for abuse monitoring, as described on our AI data use page.
• Small team. We are a small company. We keep our systems simple so that a small team can secure them well.

Questions

Security questions: security@zrico.com. Everything else: support@zrico.com.