Anthropic has Anthropic announced new capabilities that will allow its newest and largest models to end conversations when they encounter what it describes as “rare extreme cases of persistently abusive or harmful user interactions
The company is not claiming that their Claude AI models can be harmed or sentient by their conversations with other users. Anthropic says it is “highly uncertain” about the moral status of Claude or other LLMs now or in future.
Its announcement cites a recent It’s also only supposed to happen “in extreme edge cases,” like “requests for sexual content involving children and attempts to obtain information that could enable large-scale violence and acts of terrorism.”October 27-29, 2025
When Claude does end a conversation, Anthropic says users will still be able to start new conversations from the same account, and to create new branches of the troublesome conversation by editing their responses.
The company says that they are treating this feature as a continuous experiment and will continue to refine their approach.
Anthony Ha, TechCrunch weekend editor, is Anthony Ha. He worked as a reporter for Adweek, a senior at VentureBeat and a reporter for the Hollister-Free Lance. He was also vice president of content in a VC company He lives in New York City. View Bio
