Write the Line Down

On Oct. 8 Anthropic updated the rules for using Claude, and one new line got the attention: users may not "engage in sustained and needless abusive or cruel behavior toward our models." The Verge called it one of the most significant changes. It takes effect Nov. 12.

Here's why it matters beyond the AI-welfare debate. The rule is narrow on purpose. It covers only "extreme cases," and it explicitly excludes frustration, pushback, dark creative themes and testing. The penalty is mostly that Claude can end the conversation. Anthropic says it's "deeply uncertain" whether any of this matters to the model. It wrote the line anyway.

That's a values statement that actually does something. Most corporate values pages are a list of nouns: integrity, innovation, customers. This one is a sentence with a boundary in it, a reason, and a consequence. You can argue with it. That's the point.

If you want to show what you stand for without sounding like a poster in a break room, pick one specific thing you won't do. Say why in plain words. Say what happens if someone crosses it. Then live with the criticism, because you will get some, and a draft code of conduct from Microsoft's AI chief, Mustafa Suleyman, called the idea of model welfare wrong.

A line you wrote down beats a value you only claim.