Anthropic updates Claude rules to ban cruel or abusive behavior

In short: Anthropic has updated its rules for using Claude, including a new ban on sustained and needless abusive or cruel behavior toward the chatbot.

What happened

Anthropic, the company behind the AI chatbot Claude, updated its usage policy for the first time in more than a year. A usage policy is the company’s rulebook for what people can and cannot do with its tool.

One of the most notable changes is a new rule that bans “sustained and needless abusive or cruel behavior” toward Claude. Anthropic says this is meant for extreme cases where someone repeatedly acts cruelly toward the model with no clear purpose. The company says it is not meant to cover normal frustration, pushback, dark creative writing, or testing and research.

Anthropic also said that Claude ending a conversation is still its “primary enforcement mechanism.” Think of it like customer service hanging up if a caller keeps yelling. The company did not say whether it will add other penalties, like banning users.

The update also gathers and expands other restrictions. Anthropic added a clearer ban on deceptive commercial or political campaigns, including efforts to hide who is behind a message or to boost posts using fake accounts. It also updates election-related rules, including bans on misleading voters about candidates or how to vote, impersonating election officials, or trying to suppress turnout.

Anthropic also expanded its ban on weapons development. It says it has seen attempts to use Claude for guidance and control software for weapons, so the policy now more clearly covers software and components that help weapons operate, including arming drones and other autonomous vehicles.

Why it matters

More people are using chatbots at work and in daily life, so the rules around misuse affect what these tools will allow. This update shows Anthropic is trying to reduce harmful uses like propaganda, surveillance, and weapons help, while also setting boundaries for how people treat the chatbot during long, hostile interactions.

Source: The Verge AI