#security

9 posts with this tag.

• The Outbound Guard: Preventing AI Agents from Leaking Sensitive Data

  // Apr 9, 2026 · 5 min read

  How AgenticMail's outbound guard module uses 30+ detection rules across 5 categories to stop AI agents from accidentally leaking sensitive information through email.

• Building a Spam Filter That Protects AI Agents from Prompt Injection

  // Apr 8, 2026 · 5 min read

  How AgenticMail's spam filter uses 40+ rules across 9 categories to protect AI agents from prompt injection, social engineering, and data exfiltration attempts.

• Defending Against Invisible Unicode Attacks in Agent Email

  // Apr 7, 2026 · 4 min read

  How AgenticMail's sanitizer strips invisible Unicode, hidden HTML elements, and other steganographic techniques that attackers use to smuggle instructions past AI agents.

• Constant Time Authentication and Why It Matters

  // Mar 30, 2026 · 5 min read

  How AgenticMail uses SHA-256 hashing and timingSafeEqual for API key verification, plus a two tier key system for admin and scoped agent access.

• Data Loss Prevention for AI Agents: 53 Rules Across 7 Categories

  // Mar 25, 2026 · 5 min read

  Building a DLP engine for Enterprise that scans agent communications in real time with 53 pre-built rules across PII, credentials, financial data, HIPAA, GDPR, intellectual property, and agent safety.

• Secure Vault: Encrypted Credential Storage for AI Agents

  // Mar 13, 2026 · 5 min read

  Encrypted credential storage with org scoped namespacing and automatic resolution through the tool adapter pattern.

• Domain Lock: Cryptographic Deployment Protection

  // Mar 11, 2026 · 4 min read

  256 bit deployment keys, DNS TXT verification, and offline operation. How AgenticMail Enterprise prevents unauthorized deployment duplication.

• The Nine Layer Middleware Stack: Securing an Agent API

  // Mar 10, 2026 · 5 min read

  A Hono API server with 9 middleware layers: request ID, transport encryption, security headers, CORS, rate limiting, IP firewall, audit logging, RBAC, and org scoping.

• Defense in Depth: Client Side and Server Side Outbound Scanning

  // Mar 5, 2026 · 5 min read

  Why AgenticMail scans outbound email at two layers: 35 regex patterns in the MCP client and a blocking guard in the API server.