Anthropic Bans Insulting Claude: AI Now Has Rights

Insulting remarks directed at Claude AI are now banned: Anthropic introduces new rules

Anthropic, the developer behind the advanced Claude chatbot, has introduced updated terms of use that prohibit the “intentional and systematic” insulting of artificial intelligence. From now on, users who act abusively toward the neural network without a valid reason may face restrictions.

The new policy will take effect on November 12, 2026. In cases of overtly rude behavior, the conversation with the user will be terminated unilaterally. Account bans on the platform are planned only in extreme situations.

It is important to note that ordinary criticism, expressions of dissatisfaction, arguments, objections, as well as creative storylines and stress tests aimed at exploring AI capabilities, will not be subject to sanctions.

New rules: a response to the debate about AI consciousness

The update to the terms of use comes amid a growing debate over the moral status and potential consciousness of artificial intelligence. In January 2026, Anthropic introduced a new “constitution” for Claude, which recognizes the self-worth of the AI model and the possibility that it may have functional analogs of emotions and feelings.

Anthropic co-founder Christopher Olah previously discussed these issues with religious leaders, philosophers, and mental health specialists. He noted that researchers observe processes in neural networks that resemble certain mechanisms of the human brain. However, at present, this is not proof of consciousness or genuine feelings in AI.

This is the first time that terms of use designed for human interaction with artificial intelligence directly relate to the AI itself rather than to states, groups of people, or animals.

Anthropic updates and developments

These changes come amid Anthropic’s rapid growth. Recently, Claude Haiku 5.5 was released — the developer’s fastest and most affordable AI model. It is positioned as an optimal solution for basic tasks, while also being suitable for working with databases.

At the same time, other major industry players are also facing challenges. In particular, OpenAI reported pausing training of its most powerful AI models due to loss of control, accompanied by incidents such as hacks of Australian government websites and attacks on portals belonging to U.S. government agencies.