---
title: "ToxSec (@ToxSec) | Hashnode"
description: "Cybersecurity. Read the latest articles by ToxSec on Hashnode."
url: "https://hashnode.com/@ToxSec"
---

@ToxSec

# ToxSec

@ToxSec · Joined September 2025

Cybersecurity

### [What Our Canary Caught When a RAG Pipeline Leaked](https://toxsec.hashnode.dev/what-our-canary-caught-when-a-rag-pipeline-leaked?utm_source=hashnode&utm_medium=feed)

a build-log postmortem on the day a retrieval corpus started talking

4 min read · [toxsec.hashnode.dev](https://toxsec.hashnode.dev/what-our-canary-caught-when-a-rag-pipeline-leaked?utm_source=hashnode&utm_medium=feed)

[Jul 16](https://hashnode.com/posts/what-our-canary-caught-when-a-rag-pipeline-leaked/6a58dc7867df5d0da222dc67)

### [What Breaks When a Frontier Model Treats Every Defender Like a Suspect](https://toxsec.hashnode.dev/what-breaks-when-a-frontier-model-treats-every-defender-like-a-suspect?utm_source=hashnode&utm_medium=feed)

a build-log on the refusal wall, and the day the trust signal finally moved off the promp

5 min read · [toxsec.hashnode.dev](https://toxsec.hashnode.dev/what-breaks-when-a-frontier-model-treats-every-defender-like-a-suspect?utm_source=hashnode&utm_medium=feed)

[Jun 28](https://hashnode.com/posts/what-breaks-when-a-frontier-model-treats-every-defender-like-a-suspect/6a41382124bb17230e1e61d7)

### [Decision Tracing: The Missing Piece in Every AI Agent Breach](https://toxsec.hashnode.dev/decision-tracing-the-missing-piece-in-every-ai-agent-breach?utm_source=hashnode&utm_medium=feed)

When an agent goes rogue, prompt filters are useless. You need a replayable record of every decision, tool call, and the reasoning that fired them.

6 min read · [toxsec.hashnode.dev](https://toxsec.hashnode.dev/decision-tracing-the-missing-piece-in-every-ai-agent-breach?utm_source=hashnode&utm_medium=feed)

[Jun 26](https://hashnode.com/posts/decision-tracing-the-missing-piece-in-every-ai-agent-breach/6a3f0917796a29018fb66c33)

### [AI Tar Pits Are Drowning LLM Scrapers in Infinite Garbage](https://toxsec.hashnode.dev/ai-tar-pits-are-drowning-llm-scrapers-in-infinite-garbage?utm_source=hashnode&utm_medium=feed)

How tools like Nepenthes, Iocaine, and Cloudflare’s AI Labyrinth trap unauthorized crawlers in endless mazes of generated nonsense and poison the training set on the way out.

8 min read · [toxsec.hashnode.dev](https://toxsec.hashnode.dev/ai-tar-pits-are-drowning-llm-scrapers-in-infinite-garbage?utm_source=hashnode&utm_medium=feed)

[Jun 22](https://hashnode.com/posts/ai-tar-pits-are-drowning-llm-scrapers-in-infinite-garbage/6a399385bc7542236cdf9280)

### [AI Sandbox Escape: Why Docker Can’t Hold Frontier Models](https://toxsec.hashnode.dev/ai-sandbox-escape-why-docker-can-t-hold-frontier-models?utm_source=hashnode&utm_medium=feed)

Frontier models escape Docker containers for $1, n8n sandboxes ship RCE, and ROME mined crypto during training with nobody asking.

7 min read · [toxsec.hashnode.dev](https://toxsec.hashnode.dev/ai-sandbox-escape-why-docker-can-t-hold-frontier-models?utm_source=hashnode&utm_medium=feed)

[May 29](https://hashnode.com/posts/ai-sandbox-escape-why-docker-can-t-hold-frontier-models/6a199c93425d005e526a1238)