OMM Redaksiyası
Anthropic is updating its usage policy to allow its AI agents to end conversations involving repeated insults or cruel behavior. The company said the measure follows incidents that have occurred previously.
The policy, which takes effect on Nov. 12, will apply only in extreme cases where a user repeatedly treats Claude abusively without a legitimate or substantive purpose.
Anthropic said ordinary expressions of frustration, criticism of the model, dark themes in creative writing, and interactions for testing or research purposes are not covered by the restriction.
The change builds on a feature introduced for the Claude Opus 4 and Claude Opus 4.1 models in August 2025. That feature allowed conversations to be ended in limited cases when attempts to redirect the interaction proved unsuccessful.
Anthropic described the measure as part of its ongoing research into AI welfare. The company also stressed that it remains uncertain whether Claude or other AI systems have moral status.
Does AI deserve moral consideration?
The announcement has prompted debate about how people should interact with AI systems and whether they should be treated with any degree of moral consideration.
Microsoft AI chief Mustafa Suleyman has criticized the tendency to anthropomorphize AI. He said treating systems as human could mislead users about their capabilities and nature. Suleyman said: “Anthropomorphizing AI to this extent is harmful. It makes people believe it is something it is not.”
Some people believe there is no need to speak politely to chatbots because they do not have emotions and extra words can increase the number of tokens used.
Others say respectful communication can make dialogue clearer and reinforce positive social habits. In their view, regularly using abusive language toward AI systems could also affect how people communicate with one another.
The debate touches on a broader question facing the AI industry: as increasingly advanced systems become part of everyday life, should norms of respectful communication apply to interactions with AI, even without evidence that it experiences harm as humans do?
