Anthropic, the developer of the Claude AI models, on Oct. 8 published its annual updated Usage Policy, which for the first time includes a ban on “sustained and needless abusive or cruel behavior toward our models.” The new policy takes effect on Nov. 12.
The company said the rule applies only to extreme cases in which a user deliberately and repeatedly treats the model cruelly with no clear purpose, and does not cover users’ ordinary frustration, objections to answers or creative work with dark themes, nor the testing of and research on models.
Ending the conversation as the main enforcement tool
The move is in line with a feature the company introduced in August 2025 that allows Claude, in rare cases, to end conversations with users who persist in abuse on the Claude.ai and Claude Code platforms. Anthropic said Claude’s ability to end such interactions will remain the main means of enforcing the rule.
The company had presented that feature as part of its research into what is known as “model welfare,” saying at the time that it remained deeply unsure about the possible moral status of language models. There is still no scientific consensus on whether artificial intelligence systems possess any form of consciousness.
Limits on weapons and drones
The new policy makes clear that the ban on using Claude to develop weapons extends to guidance and control software and its components, as well as to arming drones and autonomous vehicles. When the models are connected to equipment capable of causing injury, it also requires that a qualified operator be able to monitor and halt them, and that the equipment return to a safe state if the connection to Claude is lost.
