FourCandles

How FourCandles Handles Your Data

A plain-English guide for business users and teams.

The short version: When you use FourCandles, your conversations stay yours. No other user can see them. No administrator can read them. Your prompts are sent to the AI provider to generate a response and to nowhere else.

What we store — and why

FourCandles saves your chat history so you can return to a previous conversation without losing your work. Your messages are stored in our database, associated with your account only.

What is stored: your messages and the AI's replies (linked to your account only); credit usage records.

What is never stored: files you attach. Documents, spreadsheets, and PDFs are processed in memory on our server to extract their text, passed to the AI, and then discarded. The original file is never written to our database or to disk.

Encryption at rest

Every message — your prompts and the AI's replies — is encrypted before it is written to the database using AES-256-GCM, the same standard used by banks and governments worldwide.

The database holds ciphertext, not plain text. The encryption key is stored separately from the database, in a secure environment secrets store. An attacker would need to compromise both simultaneously to read any message content.

This protection applies to all stored conversations across every feature of FourCandles — standard chat, Compare mode, custom AI Skills, and Projects.

Who can access your conversations

Other users: Nobody else on FourCandles can see your conversations. Every request to retrieve a session or its messages is checked against your account.

FourCandles administrators: Administrators can manage accounts and credits. There is no administrative interface to read user conversations.

Logging and analytics

Our servers log operational data — request timings and credit deductions — to keep the platform running well. We do not log the content of your prompts or AI responses. We do not use any third-party analytics service that receives your conversation data.

The one boundary that matters: AI providers

To generate a response, your message must be sent to the AI provider — OpenAI, Anthropic, Google, or Perplexity, depending on the model you choose.

FourCandles uses the API tiers of each provider. Under standard API terms of OpenAI, Anthropic, and Google, data sent via the API is not used to train their models by default.

Summary

Questions: hello@fourcandles.ai