October 10, 2026
Anthropic has provided its artificial intelligence model, “Cloud,” with a feature that allows it to automatically end conversations in specific cases of abusive or harmful interaction, as part of regulatory measures aimed at controlling dialogues that exceed the limits of acceptable use.
The company clarified, through an update to its policy, that this procedure does not apply to passing insults or the usual expression of frustration, nor does it include criticizing the model’s answers or testing its capabilities. Rather, it is limited exclusively to cases of repeated and deliberate abuse that make it impossible to continue the dialogue appropriately, to serve as a final step after attempts to redirect the discussion have failed.
It is noteworthy that “Anthropic” had begun granting some versions of “Cloud” this authority since August 2025 to deal with extreme cases. If the system resorts to ending the conversation, the user will be able to start a new conversation immediately, without this having any effect on the rest of his previous conversations or his status on the account.
