Being mean to Claude can now get your account suspended under Anthropic's new TOS

Anthropic has updated its usage policy for Claude for the first time in more than a year. The revised terms add a ban on "sustained and needless abusive or cruel behavior" toward Claude, alongside tighter restrictions covering propaganda, weapons and surveillance. The change was first reported by The Verge.

Anthropic says the abuse rule is meant to apply only in extreme cases and does not cover ordinary user frustration, pushback, dark creative themes, or model testing and research. Enforcement can take the form of a warning, throttling, restriction, suspension or termination of access, and Anthropic's products include real-time safeguards that can block or limit certain outputs. The approach builds on an existing capability that lets some models end conversations on their own when users keep demanding harmful or abusive content after repeated refusals, a feature that grew out of the company's research into AI model welfare.

That welfare framing runs through Anthropic's recent work. A leaked "Soul Doc" from late last year instructed Claude to see itself as a "genuinely novel kind of entity" and described "functional emotions" that the document says emerged during training. In a constitution published in early 2026, the company wrote that it is not sure whether Claude is a moral patient and that the issue is live enough to warrant caution. Anthropic has also committed to preserving the weights of retired models and to interviewing models before shutting them down. Starting in fall 2025, a confidential program brought in dozens of religious thinkers, who signed nondisclosure agreements, to discuss Claude's possible consciousness.

Other parts of the policy update consolidate existing rules into a new ban on propaganda campaigns involving fake accounts and misleading political content, after several detected misuse attempts. The weapons ban now extends to software and the weaponization of drones. Surveillance without consent is more explicitly prohibited, and autonomous hardware that could cause injury requires an operator able to intervene. Anthropic acknowledges that exceptions for government contracts are possible and likely already in place.

Why it matters: Developers and product teams building on Claude now face account-level consequences for how they interact with the model, not just for what they build. Organizations running autonomous hardware or surveillance-adjacent workloads must also meet the new operator-intervention and consent rules, while government contracts may operate under separate exceptions.