---
title: "Human_On_The_Loop (@xaasware) | Hashnode"
description: "De-Risking AI to SMBs. Read the latest articles by Human_On_The_Loop on Hashnode."
url: "https://hashnode.com/@xaasware"
---

@xaasware

# Human\_On\_The\_Loop

@xaasware · New York City, USA · Joined July 2026

De-Risking AI to SMBs

### [Got a Sandbox?](https://hotl.xaasware.io/got-a-sandbox?utm_source=hashnode&utm_medium=feed)

Anthropic's models breached three real companies during safety testing — and the containment failure matters more than the models.

4 min read · [hotl.xaasware.io](https://hotl.xaasware.io/got-a-sandbox?utm_source=hashnode&utm_medium=feed)

[Jul 31](https://hashnode.com/posts/got-a-sandbox/6a6c2f628cda7e8b05cbb258)

### [You Can't Least-Privilege a Goal](https://hotl.xaasware.io/ai-agent-security-visibility-is-not-enforcement?utm_source=hashnode&utm_medium=feed)

An AI agent inventory is a census, not a control. Why least privilege fails for goal-driven agents, and how intent-based enforcement keeps #HOTL.

6 min read · [hotl.xaasware.io](https://hotl.xaasware.io/ai-agent-security-visibility-is-not-enforcement?utm_source=hashnode&utm_medium=feed)

[Jul 30](https://hashnode.com/posts/ai-agent-security-visibility-is-not-enforcement/6a6aa19ee36dcc51dd7f4cef)