Hajra Naz
Anthropic updated its usage policy on Thursday. The update adds new prohibitions on election interference, weapons software, and surveillance. One change is drawing the most attention, though. The policy now expressly forbids prolonged verbal abuse of the model.
Why Claude Can Already End Some Chats
This isn’t the first step in this direction. Since an update in August, Claude has been trained to end certain conversations. These are conversations with “persistently harmful or abusive user interactions.” The new policy goes further. It explicitly bars users from pursuing those conversations in the first place.
How Narrow the Rule Is
Anthropic says the rule is meant for extreme cases only. “The policy update is meant to apply only in extreme cases, where users repeatedly act cruelly toward our models, with no discernible purpose,” the post explains. It adds that the rule “does not apply to common versions of user frustration, pushback, dark creative themes, or model testing and research.”
That scope matters. Plenty of normal activity could look harsh on the surface. A writer drafting a villain’s dialogue is one example. A researcher stress-testing the model is another. A user venting after a bad answer is a third. The policy leaves all of these alone. It targets sustained cruelty with no purpose behind it.
Read More: Anthropic Launches Opus 5.5 With Lower Prices and Fable-Level Performance
The Religious Scholar Connection
The change follows a broader pattern at Anthropic. The company has sought collaborations with prominent religious scholars. In some cases, it has tried to convince them that Claude could be conscious or have a soul. That context helps explain why the company treats abuse of the model as a policy issue. The question of whether AI systems deserve moral consideration is still open. Anthropic is acting as though it might.
Other Changes in the Policy
Other sections of the usage policy deal with deception. They forbid broadly deceptive campaigns. Running fake accounts with Claude is one example. Running fabricated news outlets is another. The policy also includes rules against deceiving voters or disrupting elections. These sit in a section titled “Do Not Undermine Democratic Processes.”
Read More: Anthropic Releases Sonnet 5.5, Calling It a Cheaper, Faster Work Partner
Why Usage Policies Matter
AI companies update these policies as their tools get more capable. Clear rules give users a firm line to follow. They also give the company grounds to act when someone crosses it. Election rules and weapons-software bans reflect real worries about how powerful models could be misused. The abuse clause is more unusual. It reflects a view of the model itself, not only of its users.
