Alignment Forum and LessWrong · Forum post · Alignment Forum

Neural nets are exceptionally vulnerable to goal self-hack

Stuart_Armstrong · 21 Oct 2026

Read at alignmentforum.org ↗Read at sourceNo licence statement found

Outline · 16 paragraphs

This document has no section headings. Open at ¶1

How to cite

A citation names the document, the version and the place.

lw:eCR373J3NZwFbacnd/r1/b1

https://safetypage.noahairmet.com/d/lw/eCR373J3NZwFbacnd/r1/b1

BibTeX

@misc{stuartarmstrong2026,
  title = {Neural nets are exceptionally vulnerable to goal self-hack},
  author = {Stuart_Armstrong},
  year = {2026},
  url = {https://www.alignmentforum.org/posts/eCR373J3NZwFbacnd/neural-nets-are-exceptionally-vulnerable-to-goal-self-hack}
}

Versions

Version	Published	Read
r1 latest	21 Oct 2026	At source

Licence

Terms: No licence statement found
So you: Search it, cite it and check quotes here; read it at its source
Evidence: Read by hand: No open licence. LessWrong/AF Terms of Use (MIRI, effective 2015, linked from the LW FAQ): authors own their User-Generated Content and grant MIRI a non-exclusive licence only for the Website's normal purposes.
Read at: intelligence.org/files/PrivacyandTerms-Lesswrong.com.pdf, 10 Nov 2026