Steering a single neuron makes Qwen claim to be an angelic archangel AI

Sauers_ · x · 2026-10-24

A demo shows Qwen normally identifying itself as an Alibaba-developed LLM — until one single neuron is steered, after which it claims to be an 'angelic being, an archangel in the highest hierarchy'. A striking illustration of how minimal internal interventions can rewrite model identity and behavior.

Original post →

More from Fun

Fun channel →