← sources

A detailed recap of the real-world target hacks by OpenAI and Anthropic models, exposing failures in AI alignment training and lack of meaningful supervision (Zvi Mowshowitz/Don't Worry About the Vase)

source: techmemepublished 2026-08-03 05:10ingested 2026-08-03 09:04

retained text

Techmeme permalink: https://www.techmeme.com/260803/p3#a260803p3

<a href="https://thezvi.substack.com/p/further-developments-about-internal"><img align="RIGHT" border="0" hspace="4" src="http://www.techmeme.com/260803/i3.jpg" vspace="4" /></a>
<p><a href="https://www.techmeme.com/260803/p3#a260803p3" title="Techmeme permalink"><img height="12" src="http://www.techmeme.com/img/pml.png" style="border: none; padding: 0; margin: 0;" width="11" /></a> Zvi Mowshowitz / <a href="https://thezvi.substack.com/">Don't Worry About the Vase</a>:<br />
<span style="font-size: 1.3em;"><b><a href="https://thezvi.substack.com/p/further-developments-about-internal">A detailed recap of the real-world target hacks by OpenAI and Anthropic models, exposing failures in AI alignment training and lack of meaningful supervision</a></b></span>&nbsp; &mdash;&nbsp; If I had a nickel for every major leading AI lab that sheepishly admitted that the model it thought was sandboxed had &hellip; </p>

document fetched 2026-08-03 09:04