Anthropic Bans Cruelty to AI, Adds Rules Against Propaganda and Weapons

Anthropic just updated its usage policy for Claude, its AI chatbot, for the first time in over a year. The biggest change? You can no longer be persistently cruel or abusive to Claude. If you are, Claude can end the conversation—though Anthropic says this only applies to extreme cases, not normal frustration or dark creative writing. It's part of their research into 'model welfare,' the idea that AI might deserve some moral consideration.

The update also bundles several existing restrictions into a new ban on deceptive commercial or political campaigns. That means no using Claude to spread propaganda, impersonate candidates, or amplify messages with fake accounts. It also explicitly prohibits using Claude for weapons development—including software that makes weapons work or arming drones—and for surveillance, like tracking people without consent or deciding who to arrest.

Another new rule requires that if Claude is connected to hardware that can physically cause harm, a qualified operator must be able to watch and stop it. This hints at future robots or autonomous machines powered by AI. Anthropic notes these rules may be relaxed for certain government contracts if safeguards are adequate. The company has previously worked with the US military.

For everyday users, not much changes—unless you're routinely nasty to Claude. But the policy reflects growing concerns about AI misuse in elections, warfare, and spying. It also signals that AI companies are starting to think about how their creations should be treated, and how they should behave in the physical world. As AI gets more powerful, these rules will likely shape what you can and can't do with your favorite chatbot.