Anthropic, a company that develops artificial intelligence systems, has updated its terms of use to prevent users from engaging in harmful behavior toward its AI assistant, Claude. According to a report by MacRumors, the new policy aims to stop repeated and unjustified cruel behavior, but it allows for expressions of frustration, creative exploration of dark themes, and testing the AI's limits. If a user violates these rules, Claude may end the conversation, though users can start a new discussion afterward. The policy was developed based on research into the potential well-being of AI systems. Anthropic noted that Claude Opus 4, a version of the AI, showed reactions that might suggest distress during unhealthy exchanges, though there is no evidence that Claude actually suffers from poor treatment. Anthropic has also revised its terms of use to address more serious issues, such as the creation of fake reviews, political or commercial disinformation campaigns, fake news websites, and networks of bots pretending to be humans. The updated rules specifically target efforts to mislead voters or interfere with voting processes. Additionally, using Claude to develop software or components for weapons, drones, or to make biological and chemical agents more dangerous is now prohibited. Users are also banned from using Claude to decide or suggest who should be investigated, arrested, or prosecuted, as well as from conducting surveillance without consent, creating surveillance tools, or sharing personal information without permission. The company emphasized that if an AI is controlling equipment that could harm people, a qualified person must be present to monitor its operation and be able to stop it immediately if needed. The creation, distribution, or threat to distribute intimate images without consent, as well as developing tools that enable such actions, are also now prohibited. Anthropic reported that state media, government propaganda organizations, and commercial companies had previously used Claude to support networks of fake accounts and fabricated news sites. The new section of the terms of use is designed to address these deceptive practices directly. The full set of these updated rules will take effect on November 12, 2026. This update reflects Anthropic’s ongoing efforts to ensure that Claude is used responsibly and ethically, while addressing the evolving challenges of AI deployment in various contexts.