Anthropic has announced new capabilities that will allow some of its newest, largest models to end conversations in what the company describes as “rare, extreme cases of persistently harmful or abusive user interactions.” Strikingly, Anthropic says it’s doing this not to protect the…