Company Anthropic has introduced an important change to how Claude AI works: the artificial intelligence will now automatically end conversations with users who show persistent abusive or harmful behavior. This applies to all users, including those who use Claude via API.
The system does not stop at filtering individual phrases, but fully breaks off the dialogue when violations are repeated. Such behavior is classified as "harmful interaction" — for example, regular threats, insults, or attempts to use AI for bullying.
Such measures are especially important in the context of global discussions about bullying and toxicity in the digital environment. Anthropic became the first company in the AI market to introduce such a decisive automated response. If systems before could only warn the user or block individual messages, now Claude literally "leaves the chat," not allowing a toxic dialogue to continue.
This is not only a step toward protecting digital boundaries, but also a signal to society — even neural networks choose a culture of communication.
Source: The Verge
