Protects AI agents from prompt injection attacks hidden in emails by acting as a secure email gateway.
Sponsored
AI assistants with email access are highly vulnerable to prompt injection attacks, where malicious emails can contain hidden instructions that compromise data. Read No Evil acts as a crucial secure email gateway, sitting between your email provider and your AI agent, scanning every incoming email for these insidious attacks using ML-based detection before the AI processes them. This ensures that your AI agents only interact with safe, sanitized content, preventing unwanted actions and data breaches while allowing controlled email functionality.
Key Features
010 GitHub stars
02Support for configuring multiple IMAP email accounts
03ML-powered prompt injection detection using DeBERTa model
04Fine-grained, per-account permissions (read-only by default)
05Exposes email functionality via Model Context Protocol (MCP)
06Local inference with CPU-only PyTorch for efficiency
Use Cases
01Protecting AI agents from email-borne prompt injection attacks
02Implementing fine-grained access control for AI models interacting with email accounts
03Enabling secure and controlled email access for AI assistants
