---
title: "Discussion on \"AI Coding Is the New\nSoftware Supply Chain Risk\" | Hashnode"
description: "Discussion on \"AI Coding Is the New\nSoftware Supply Chain Risk\". The question is no longer \"Is AI dangerous?\" — it's \"Who verifies the code AI writes, and how?\"\n\nThe Wrong Question\nDiscussions around AI coding often get stuck on one point:\n\"Is AI dangerous?\"\nBut th"
url: "https://hashnode.com/posts/ai-coding-is-the-new-software-supply-chain-risk/69df043991716f3cfb9b42b4"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [AI Coding Is the New Software Supply Chain Risk](https://leeh.io/ai-coding-is-the-new-software-supply-chain-risk?utm_source=hashnode&utm_medium=feed)

The question is no longer "Is AI dangerous?" — it's "Who verifies the code AI writes, and how?" The Wrong Question Discussions around AI coding often get stuck on one point: "Is AI dangerous?" But th

Apr 15 · 6 min read · leeh.io  
Read the article

-   [#AI](https://hashnode.com/tag/AI)
-   [#vibe-coding](https://hashnode.com/tag/vibe-coding)
-   [#mcp](https://hashnode.com/tag/mcp)
-   [#vibecoding](https://hashnode.com/tag/vibecoding)
-   [#claude-code](https://hashnode.com/tag/claude-code)
-   [#cursor](https://hashnode.com/tag/cursor)

## Written by

[leeh](https://hashnode.com/@leeh)

@leeh · Seoul, South Korea

  
[View profile](https://hashnode.com/@leeh)

​#OpenSource #AI #Security #Python

One-way secure push pipeline for AI-generated code. Use Claude Code, Gemini CLI, Copilot freely — deliver results to your internal network safely.

## Responses 2

Markdown supported · ⌘↵ to post

[CapeStart](https://hashnode.com/@capestart)

[Apr 15](https://hashnode.com/posts/ai-coding-is-the-new-software-supply-chain-risk/69df043991716f3cfb9b42b4/comment/69df8c5870e165e0d8a8bfa0)

Would be interesting to see how many orgs actually have an inbound control layer today vs relying purely on reviews and scanners.

[leeh](https://hashnode.com/@leeh)

[Apr 17](https://hashnode.com/posts/ai-coding-is-the-new-software-supply-chain-risk/69df043991716f3cfb9b42b4/comment/69df8c5870e165e0d8a8bfa0#reply-69e1aaacf27b011bda8dca2e)

In reality, more than 90% of orgs are still stuck with traditional reviews and scanners. Finding a place that has built a real inbound control layer is like finding a needle in a haystack.

​It mostly comes down to the trade-off between security and speed. True inbound control means having your organization's policies (Policy-as-Code) integrated right at the prompt level, which is technically demanding to customize. We're likely heading toward a future where 'security proxies' sit between the IDE and the AI, filtering risks before the code even hits the editor.
