Anthropic updated its usage policy on Thursday to codify new prohibitions on election interference, weapons software, and surveillance. The most attention-getting change is a new prohibition against prolonged verbal abuse of the model.
Since an update in August, Claude has been trained to end conversations with persistently harmful or abusive user interactions. The new policy explicitly bars users from pursuing those conversations.
Policy targets extreme abuse cases
The policy update is meant to apply only in extreme cases. Users must repeatedly act cruelly toward models with no discernible purpose for the rule to trigger.
It does not apply to common versions of user frustration, pushback, dark creative themes, or model testing and research.
Read nextGitHub updates secret detection model and previews push protection checksCompany previously consulted scholars
The change comes after Anthropic has sought out collaborations with prominent religious scholars. In some cases, the company sought to convince them that Claude could be conscious or have a soul.
Separate sections of the usage policy forbid users from engaging in broadly deceptive campaigns. This includes using Claude to run fake accounts or fabricated news outlets.
Democratic processes receive protections
The policy also includes rules against deceiving voters or disrupting elections. These prohibitions are collected in a section titled Do Not Undermine Democratic Processes.
