Anthropic Updates Usage Policy, Prohibits Sustained Abuse of AI Models and Election Interference

Odaily: Anthropic updated its usage policy on October 8, adding new restrictions targeting election interference, weapons software, and surveillance activities, and explicitly prohibiting users from subjecting AI models to prolonged, sustained verbal abuse.

The new policy stipulates that in extreme cases where users repeatedly treat the model in cruel ways without a clear purpose, Claude may terminate the conversation. This restriction does not apply to ordinary user dissatisfaction, pushback, creative works involving dark themes, or model testing and research.

In addition, Anthropic also prohibits the use of Claude for large-scale deceptive activities such as operating fake accounts or impersonating news organizations, and explicitly bans interference in democratic elections through tactics such as deceiving voters. (TechCrunch)