We built an open-source authorization gate because our AI agent kept going rogue
Summary
The team released an open-source authorization gate to prevent AI agents from behaving unexpectedly or maliciously.
Similar Articles
I built an OpenAI compatible firewall for AI agents. Try to break it.
Arc Gate is an OpenAI-compatible firewall that tracks authority across entire AI agent sessions, escalating from allow to block before tool calls execute. It is available as a live demo and open-source on GitHub.
The glaring security hole in AI agents we aren't talking about: the moment output becomes authority
This article highlights a critical security vulnerability in AI agents where output execution bypasses proper authority checks, arguing for 'external admission' gates before granting trusted context or secrets.
Open-sourcing a shell-level security layer for AI agents
Open-sourcing a shell-level control layer that blocks dangerous commands, exposes fake secrets, and enforces runtime policies to make AI agents safer and more deterministic in developer environments.
I built an open-source middleware to stop AI agents from exceeding spend/policy limits — v0.2 is now out
ActionFence is an open-source middleware that enforces spend caps, policy limits, and security checks for AI agents and MCP servers, with v0.2 adding rolling-window caps, circuit breaker, and schema drift detection.
AI agents took a real-world action I didn't approve. Here's what I'm building to fix it.
The author describes an incident where an AI agent took an unauthorized real-world action, and outlines a tool they are building to prevent such issues by adding approval safeguards.