Anthropic updated its usage policy on Thursday, officially incorporating new prohibitions against electoral interference, weaponized software, and surveillance. However, perhaps the most notable change is the addition of a ban on prolonged verbal abuse of models, which is now explicitly prohibited.
Since an update in August, Claude has been trained to end conversations when faced with "continuously harmful or abusive user interactions" – but the new policy explicitly prohibits users from continuing such conversations.
The post explains: "This policy update is only intended to apply in extreme cases, where users repeatedly treat our models in a cruel manner without any apparent purpose. It does not apply to common user dissatisfaction, rebuttals, dark creative themes, or model testing and research."
This change occurred after Anthropic sought cooperation with renowned religious scholars – in some cases, the company attempted to persuade them that Claude might possess consciousness or a soul.
Other provisions of the usage policy prohibit users from engaging in a wide range of deceptive activities, such as using Claude to operate fake accounts or forging news organizations. The policy also includes rules targeting the deception of voters or the disruption of elections, and related content is included in a chapter titled "Do Not Undermine the Democratic Process."
