Anthropic updates usage policy to prohibit election interference and prolonged model abuse

Anthropic updated its 2026 usage policy on October 8, adding election interference, deceptive campaigns, weapons software and surveillance to the list of prohibited uses. It also prohibits repeated cruel or abusive interactions with models when there is no discernible purpose. Anthropic said the restriction does not target ordinary frustration, pushback, dark creative work, model testing or research. Since an August update, Claude has been able to end conversations that continue harmful or abusive interactions, and the new policy makes it clear that users may not intentionally seek those interactions. Election related rules appear in a section focused on not undermining democratic processes, including deception of voters or disruption of elections. The report did not identify one uniform effective date for every clause. The update tightens permitted use cases without changing model pricing or API capabilities.