SciHamster comments on Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMs

[SciHamster]

fwiw, the fact that somebody can just finetune the model, is already indicative of a serious problem