Anthropic updates Claude rules to ban election interference and model abuse
Anthropic updated its usage policy to explicitly prohibit election interference, weapons software, surveillance, and deceptive campaigns. The policy also bars users from repeatedly abusing Claude in extreme cases, while allowing ordinary frustration, criticism, and research. The change follows Anthropic's efforts to engage religious scholars about whether Claude could be conscious.
Anthropic's revised rules add explicit bans on using Claude for election disruption, weapons code, surveillance, and coordinated deception. They also target fake accounts and fabricated news outlets, with voter deception and electoral disruption grouped under a section on protecting democratic processes.
The policy addresses repeated cruelty toward the model only in severe cases, saying it does not cover everyday annoyance, disagreement, somber creative projects, or technical evaluation and study. This follows an August training change that let Claude end chats marked by sustained mistreatment or hostility, and comes amid Anthropic's outreach to religious scholars about Claude's possible consciousness or soul.
The update could affect developers, researchers, and political organizations that rely on Claude, potentially narrowing certain uses while leaving ordinary criticism and creative work intact. It may also encourage other AI companies to adopt clearer limits, though enforcement and definitions of extreme abuse may remain difficult to apply consistently.