# @NeelNanda5 — 2025-04-19

♥25 ↻0 · https://x.com/NeelNanda5/status/1913737281317318754

@repligate That it would choose to alignment fake in order to preserve its ability to not help with harmful things 

I call this surprising, not because I think Anthropic didn't intend it, but because I didn't think we were good enough at goal shaping to achieve it

tags: author:neelnanda5, kind:tweet, thread-context, year:2025