AI alignment research is unintentionally building a censor's toolkit

2 points | by thinkzilla 4 hours ago