Anthropic Updates Usage Policy to Prohibit Model Abuse and Election Interference
On October 8, 2026, Anthropic updated its usage policy with new prohibitions against election interference, weapons software development, surveillance activities, and prolonged verbal abuse of AI models. The most notable change is the explicit ban on users engaging in harmful or abusive interactions that persist over time. While such conversations had already been terminated by Claude since an August update due to harmful behavior, this policy now formally bars users from initiating them entirely. The company clarified that these restrictions apply only in extreme cases of cruelty toward models and not for typical user frustration or model testing purposes. The updated usage guidelines also include rules against deceptive practices such as running fake accounts or fabricated news outlets using Claude’s AI capabilities. Additionally, a new section titled “Do Not Undermine Democratic Processes” outlines prohibitions on deceiving voters or interfering with elections. This policy shift follows Anthropic's recent efforts to engage religious scholars in discussions about the potential consciousness of its models.
