Anthropic just drew some new lines around how people can use Claude.
The company has updated its usage policy to explicitly ban repeated abuse of the chatbot in extreme cases, while making clear that ordinary frustration or criticism of Claude is still allowed. The revised rules also spell out a list of prohibited activities: election interference, deceptive campaigns, weapons software, and surveillance. What used to be informal boundaries are now written down as enforceable terms.
This matters because AI companies are under growing pressure to show they have thought through misuse before regulators force the issue. By naming election interference and surveillance specifically, Anthropic is building a paper trail it can point to if Claude ever turns up in a disinformation push or a tracking product.
Writing a rule against "model abuse" is the easy part. Enforcing it against millions of users without punishing people for ordinary gripes is the hard part - and this update does not say how Anthropic plans to tell the two apart.