---
format: "aidr-story-markdown/v1"
id: "2205158e0f4925c776681dca7b150907bcff71217284ec0ab13b1188b0314b3c"
canonical_url: "https://aidr.today/2205158e?lang=en"
title: "Ask a model if code is malicious and it reaches for its morals"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-10-06T17:20:26.000Z"
category: "Research"
topics: ["safety"]
source_urls: ["https://www.manifold.security/blog/do-models-consider-morality-malware","https://news.ycombinator.com/item?id=49981455"]
summary: "When asked whether code is malicious, a mixture-of-experts model recruits a morality-associated expert path."
---

# Ask a model if code is malicious and it reaches for its morals

> [Open the canonical story](<https://aidr.today/2205158e?lang=en>)

**Published:** 2026-10-06T17:20:26.000Z
**Category:** Research
**Topics:** safety

## Summary

When asked whether code is malicious, a mixture\-of\-experts model recruits a morality\-associated expert path\.

## Sources

- [Story source](<https://www.manifold.security/blog/do-models-consider-morality-malware>)
- [Discussion](<https://news.ycombinator.com/item?id=49981455>)