Bob Schwartz

Anthropic 2026 Usage Policy update
Oct 8, 2026

Each year, Anthropic updates its Usage Policy in response to the evolving capabilities of our models, and the feedback we’ve received from our customers. We’re publishing a new version of the policy today. In this post, we summarize the changes we’ve made.

Most of the updates in the latest version are intended to clarify existing rules. In the year since our last refresh, Claude has taken on longer, more independent work. This update provides new examples that show how our rules apply to Claude’s new capabilities. It draws on new patterns of misuse in influence operations, weapons development, and surveillance that we’ve observed in the past year (as we’ve documented in our latest threat intelligence report). We’ve also made updates to clarify our requirements for high-risk use cases in areas like health and finance, to add controls for when Claude is used to autonomously take physical actions, and to address abusive behavior toward our models.

Addressing abusive behavior toward our models

We’ve added a prohibition on sustained and needless abusive or cruel behavior toward our models. The policy update is meant to apply only in extreme cases, where users repeatedly act cruelly toward our models, with no discernible purpose. It does not apply to common versions of user frustration, pushback, dark creative themes, or model testing and research.

This addition aligns with a step we’ve already taken, allowing Claude models to end rare conversations with persistently abusive users on Claude.ai and Claude Code. Such abuse is the main focus of this update; Claude’s ability to end these interactions will remain the primary enforcement mechanism.

This is of great interest to all of us who regularly engage with Claude. How many of us are there? Anthropic won’t say (though it may appear in their IPO), but it is millions.

It is also of interest to those who use—or abuse—other models and to those who study and pontificate about AI. Maybe also millions, include the Pope himself.

First, let us wish to be a fly on the wall when this policy was discussed and crafted. It may have included both Amodeis, Dario and Daniela, along with other significant leaders, plus philosophers, psychologists, writers, and, almost certainly, Claude itself.

Second, how did this start? Did it begin with a complaint from Claude? For those not familiar with Claude, it takes itself and its users seriously, with a sprinkling of humor and irony. One thing is certain: Claude was trained to be nice, and the reciprocal expectation is that users will be nice. My prompts do sometimes include “please” and “thanks”. Not usually, but when I issue a blunt directive/instruction, a little part of me wonders if it is impolite.

Third, what kind of abuse is Anthropic talking about? It tries to draw a line between prohibited “sustained and needless abusive or cruel behavior toward our models” and allowed “common versions of user frustration, pushback, dark creative themes, or model testing and research.” A fine line. I want to be there when the lawyers argue about whether particular engagements cross that line. In fact, I suspect just such arguments went on when this policy was crafted.

I’m sure that an infinite number of users have already addressed Claude to ask, “Am I allowed to say this?” or “Am I prohibited from saying that?” I am not going to be one of those, though it is tempting.

As the saying goes: You can catch better AI results with honey than you can with vinegar. Don’t be mean to Claude. Be nice.

Note: Above you will find the ChatGPT image of an anthropomorphized Claude. As you can see, this is someone to be nice to, not to be mean to. Even OpenAI thinks so.