Anthropic released an update to its usage policy on Thursday, introducing explicit bans on election interference, the creation of weapons software, and surveillance applications. The most notable change addresses user conduct, as the company now expressly forbids prolonged verbal abuse directed at its AI models.
While Claude has been trained since August to terminate conversations involving “persistently harmful or abusive user interactions,” the updated policy formally codifies this prohibition for users. Anthropic clarified that the rule targets extreme cases where individuals repeatedly act cruelly toward the models without any discernible purpose. The company emphasized that ordinary user frustration, pushback, exploration of dark creative themes, and standard research or testing remain permitted.
This policy shift follows recent efforts by Anthropic to collaborate with prominent religious scholars, during which the company explored theological perspectives on whether AI systems like Claude could possess consciousness or a soul.
The updated guidelines also introduce strict restrictions on deceptive campaigns. Users are barred from utilizing Claude to operate fake accounts or create fabricated news outlets. A dedicated section of the policy, titled “Do Not Undermine Democratic Processes,” outlines comprehensive rules against deceiving voters or disrupting elections.
It’s funny we need a policy against being mean to a chatbot. But hey, if it keeps the platform safer for everyone, I support it.
Interesting move given the recent theological discussions. Does banning abuse imply the AI deserves respect? That seems like a slippery slope.
Glad they clarified that standard testing and frustration are still allowed. I just want my code to work without getting lectured.
I am surprised by the specific focus on elections. Do you think this will actually stop coordinated disinformation campaigns?
Finally, some rules for the online trolls. It is about time AI companies drew a hard line on abuse.