AI Safety Radar
Tuesday, 10 November 2026

AI-safety startups

Filters

Clear all

October 2026

1
AI safetyShow HN253 points18 Oct

Sphinx

A hook that quizzes developers on AI-written code before they push to production.

September 2026

2
AI safetyShow HN185 points1 Sept

Weedout

Safari extension for Mac that hides YouTube videos labeled "Made with AI" from feeds, search and Shorts.

July 2026

3
AI safetyShow HN115 points28 Jul

Formally verified 3D CSG

Mesh intersection code for 3D solid modeling, proven correct in Lean 4 so reviewers need only read the spec.

AI safetyShow HN110 points23 Jul

OneCLI

Open-source gateway that keeps API keys away from AI agents and enforces request policies for developers running them.

AI safetyShow HN226 points13 Jul

Clawk

Disposable Linux VMs that let developers run coding agents away from their laptops.

June 2026

2
AI safetyShow HN112 points9 Jun

Claw Patrol

Network firewall that lets AI agents reach production systems while humans and LLMs approve destructive actions.

May 2026

2
AI safetyShow HN687 points19 May

Forge

Open-source guardrails that make small self-hosted LLMs reliable at multi-step tool-calling workflows for local users.

AI safetyShow HN129 points8 May

Git for AI Agents

Open-source history tool for Claude Code sessions that records what the agent changed and why.

April 2026

1
AI safetyShow HN156 points22 Apr

Agent Vault

Open-source proxy that holds credentials so AI agents can call services without ever seeing the secrets.

March 2026

3
AI safetyShow HN141 points30 Mar

Zerobox

Command-line sandbox that blocks file and network access and hides API keys from commands, for developers running AI agents locally.

AI safetyShow HN161 points12 Mar

OneCLI

Open-source proxy that gives AI agents placeholder keys and swaps in real credentials, so developers' agents never see secrets.

February 2026

3
AI safetyShow HN201 points24 Feb

enveil

Tool that hides .env secrets from AI coding assistants so developers' API keys stay out of their context.

AI safetyShow HN328 points24 Feb

Steerling-8B

An 8B language model that can explain any token it generates, according to its one-line listing.

January 2026

2
AI safetyShow HN146 points30 Jan

Amla Sandbox

Sandbox that runs AI-generated code in a WebAssembly bash shell, limited to tools developers provide.

AI safetyShow HN122 points12 Jan

Yolobox

Sandbox letting developers run AI coding agents with sudo without risking their home directory.

December 2025

1