• Services
  • Blog
  • Join
  • Contact

Writing

Blog

Analysis, techniques, and perspectives from the disreGUARD team on prompt injection and AI security.

February 7, 2026

The auditor in the airlock: a security pattern for AI agent decisions

When an agent needs to make a security-sensitive decision about tainted data, you need an information bottleneck between the taint and the judgment. Here's how to build one.

February 7, 2026

Hardening OpenClaw: a practical prompt injection defense

This week researchers demonstrated persistent backdoors in OpenClaw via prompt injection. We're helping harden it with `sig` and patterns apply to any agent framework.

February 6, 2026

sig: instruction signing for prompt injection defense

We can create a clear trust boundary by signing instructions and giving models a tool to participate in making secure choices

February 5, 2026

Injection is inevitable. Disaster is optional.

Prompt injection is an infrastructure problem. We can't prevent it, but we can massively reduce the risk and impact.

  • Services
  • Blog
  • Newsletter
  • Join
  • X
  • Bluesky
  • Instagram

© 2025 disreGUARD. All rights reserved.