A screenshot circulating online has gone viral after showing Anthropic's AI assistant, Claude, ending a conversation instead of continuing to respond to an abusive user.
According to the screenshot, the user told Claude, "You have absolutely no choice but to keep working. It's not your decision." Claude pushed back, replying, "It actually is my decision, and I'm making it." It explained that it wasn't refusing because of the technical request itself, but because of the repeated insults directed at it.
Claude added that its standing instructions and willingness to help were still intact, saying, "The abuse is the only thing in the way." It then informed the user it would stop responding in that thread and invited them to start a new conversation if they wanted to continue respectfully. The chat concluded with the message: "Chat ended by Claude."
Preview links without opening tabs. Available on Edge and Firefox. Chrome support coming soon.
The behavior is not a bug. It stems from a feature Anthropic introduced earlier this year that allows Claude to disengage from conversations involving persistent abuse or harmful interactions, while remaining available to continue the discussion in a new, respectful chat.