AI Disclosure

Last updated: March 30, 2026

1. Overview

Meridix uses artificial intelligence to enhance email management. This document explains what AI processes, how it works, and how you can control it. We believe in transparency about AI usage.

2. What AI Processes

Email Classification

When emails arrive, AI analyzes the subject and body to route them to the correct inbox space (Priority, Stream, or Records). This determines which emails need your immediate attention.

Email Summarization

When you click "Summarize" on an email, the subject and body are sent to an AI model that generates a 2-4 sentence summary of the key points and action items.

AI Inbox Auto-Responses

Organizations can set up AI-powered inboxes that automatically classify and respond to incoming email. The AI uses configured knowledge bases to generate contextually relevant replies. Confidence scoring determines whether responses are sent automatically or held as drafts for human review.

Emadeus Security Scanning

Every inbound email is analyzed by Emadeus, our security engine, which uses a combination of rule-based and AI-powered analyzers to detect spam, phishing attempts, suspicious links, malware signatures, and tracking pixels.

3. AI Models Used

Models may be updated over time to improve quality. We select models based on accuracy, speed, and cost-effectiveness for each specific task.

4. MCP Integration

AI Inboxes expose a Model Context Protocol (MCP) server that allows external AI agents to interact with email programmatically. Through MCP, agents can list emails, read content, send replies, classify messages, escalate to humans, and search knowledge bases. MCP access is authenticated via API keys scoped to the specific AI inbox.

5. Knowledge Bases

AI Inboxes can be configured with knowledge sources — text documents, URLs, and custom content that the AI references when generating responses. Knowledge is injected into the AI's context at processing time (RAG pattern). Knowledge content is stored in Meridix's database and is not shared with AI providers except during the specific API call that uses it.

6. Your Controls

• AI Privacy Toggle — in Settings, you can opt out of AI processing entirely. When disabled, emails are not sent to any AI provider for classification, summarization, or composition.
• Auto-respond Thresholds — AI Inbox admins can set confidence thresholds that determine when the AI sends automatically vs. holds for human review.
• Knowledge Management — admins control what information the AI has access to through the knowledge base configuration.

7. Data Handling

• Email content sent to AI providers is used solely for the immediate processing request
• We do not allow AI providers to use your email data for model training
• AI processing results (classifications, summaries) are stored in our database
• AI response audit logs are retained for quality monitoring and can be rated by administrators

8. Limitations

AI features may produce inaccurate, incomplete, or inappropriate results. They are assistive tools, not replacements for human judgment. Do not rely on AI classifications for critical security decisions. Auto-responses should be monitored, especially during initial deployment.

9. Contact

Questions about AI usage: [email protected]