Claude Ai Chatbot from Antarbur now can now end conversations that are considered “constantly harmful or offensive”, as monitored Earlier Techcrunch. ability Available in OPUS 4 and 4.1 modelsAnd Chatbot will be allowed to end the conversations as the “last resort” after users have repeatedly requested to create harmful content despite rejection and multiple attempts to re -direct. The aim of this is to help “potential luxury” for artificial intelligence models, as Antarubor says, by ending the types of reactions in which Claude showed “clear distress.”
If Claude chooses a conversation cut, users will not be able to send new messages in that conversation. They can still create new chats, in addition to editing and re -attempting to follow if they want to follow a specific topic.
During her test for Claude OPUS 4, Anthropor says she found that Claude has a “strong and harmonious aversion”, including when they are asked to generate sexual content that includes the palace, or providing information that can contribute to violence and terrorism. In these cases, a person says that Claude has shown “a pattern of apparent distress” and “the tendency to end the harmful conversations when giving ability.”
Anthropor notes that talks that lead to this type of response are “extremist edge cases”, adding that most users will not face this road barrier even when chatting on controversial topics. AI released the Claud starting not to finish conversations if the user shows signs that they might want to harm themselves or cause “imminent damage” to others. Humanitarian partners With hybridization lines, online crisis support, to help develop responses to claims of self -harm and mental health.
Last week, Antarbur has also updated the policy of using Claude because artificial intelligence models are more rapidly over safety. Now, the company prevents people from using Claude to develop biological, nuclear, chemical or radiological weapons, as well as to develop a harmful symbol or exploit the weaknesses in the network.