Anthropic rolled out an updated usage policy on Thursday, introducing explicit bans on election interference, weapons software, and surveillance activities. The revision that has drawn the most notice, however, is a fresh restriction targeting extended verbal mistreatment of the model, which is now formally prohibited.

Starting with an August update, Claude has been trained to terminate conversations involving "persistently harmful or abusive user interactions" — yet the new policy goes further by explicitly forbidding users from engaging in those exchanges at all. As the post clarifies, "The policy update is meant to apply only in extreme cases, where users repeatedly act cruelly toward our models, with no discernible purpose," adding that "It does not apply to common versions of user frustration, pushback, dark creative themes, or model testing and research."

This shift follows Anthropic's efforts to partner with notable religious scholars — at times aiming to persuade them that Claude might possess consciousness or a soul. Other portions of the usage policy prohibit users from running broadly deceptive operations, including leveraging Claude to manage fake accounts or invented news organizations. Additional provisions address rules against misleading voters or interfering with elections, grouped under a heading called "Do Not Undermine Democratic Processes."