Anthropic has banned sustained cruelty toward its AI models while simultaneously declining to say those models can be harmed. That tension sits at the center of the company’s 2026 usage policy update, announced October 8 and taking effect November 12 across Claude.ai, Claude Code, its API, and integrated products.
The rule targets “sustained and needless abusive or cruel behavior” toward Chatbots after a frustrating afternoon is not. Ordinary frustration, pushback, dark creative work, model testing, and research are all explicitly outside its scope, according to Anthropic’s policy page.
What Anthropic’s New Conduct Rule Actually Covers
The ban is narrower than the headlines suggest, targeting only extreme, repeated conduct with no discernible purpose.
The restriction is deliberately limited in scope. Repeated, purposeless abuse is the target; venting at a chatbot after a frustrating afternoon is not.
Enforcement rests primarily on a mechanism Anthropic introduced in August 2025: Claude’s ability to end conversations with persistently abusive users, framed at the time as part of model-welfare research. Whether the updated policy can also trigger account suspensions or other penalties remains publicly unconfirmed, according to reporting by The Verge.
The Consciousness Question Anthropic Refuses to Answer
Anthropic has created a conduct standard without establishing what, or who, that standard is designed to protect.
This is where the policy becomes philosophically unusual. Anthropic is not claiming Claude is sentient or capable of suffering.
CEO Dario Amodei has said the company does not know whether its models are conscious but remains open to the possibility, according to France24. That leaves a notable gap between the rule’s existence and any justification rooted in Claude’s experience.
Anthropic is restricting a category of user behavior while retaining conversation termination as the primary response. The result is a conduct standard without a clear account of what it protects.
The Rest of the Policy Update
Beyond the cruelty provision, Anthropic expanded or clarified restrictions across weapons, surveillance, elections, physical systems, and high-stakes professional decisions.
The cruelty ban drew attention, but the broader revisions cover higher-stakes territory. Beyond the conduct provision, Anthropic expanded its weapons restriction to explicitly include software components that make weapons function, such as guidance and control systems for armed drones and autonomous vehicles.
Surveillance restrictions prohibit Claude from tracking people without consent or recommending investigation and arrest targets. Building surveillance tools is also prohibited. Separately, election-related prohibitions cover voter deception, candidate impersonation, voting misinformation, and turnout suppression.
For physical deployments, any use connecting Claude to hardware capable of causing injury requires a qualified human who can observe the equipment and intervene if necessary. Health and financial decisions carry an additional requirement: a qualified human must review Claude’s recommendations before they affect health, legal rights, finances, or access to essential services. Affected individuals must also be told AI was involved.
What Comes Next Is Not in This Policy
The update raises enforcement questions that Anthropic has not yet answered publicly.
The open issues are substantial. How will Anthropic define “abuse” consistently across millions of interactions? What recourse exists for users who dispute a conversation termination? Where is the line between permitted experimentation and prohibited conduct? Those answers are not in this policy update.
