My Claude Code just puts it in the commit which anyone can read before pushing. Is that not the case here?

I don't use git features in vscode, but from what I understand the user clicked some button to make a commit, typed in a commit message, and then hit "OK" and the editor called `git commit ...` in the background... after silently adding "Co-Authored by Microsoft Copilot" to the commit message.

That's a little different than Claude doing the commits all by itself and happening to include an attribution line. Especially since, as it turns out, this was being done on clients that had all the AI stuff turned off. But even if that weren't the case, it'd still be wrong.

Also you shouldn't be using Claude that way...

It's technically the same thing because a pre-commit hook can easily remove it.

I did this with the very first versions of claude which didn't have a documented setting to turn it off, and kept it every since. It works with every single coding tool because it just looks for the same key word.

Nothing wrong at all with separating out Claude’s work with commits! In fact, it’s preferable IMO — it lets people browsing the history identify code that was primarily written by AI.

But it actually doesn't.

This is not just a hypothetical but a non-common workflow: I already wrote upstaged code change myself. I ask claude to review it, and if ok, commit and push.

At no point did claude author any of it, just a review. So a co-author statement is false.

Anthropic documents how to turn it off, though, and shows you in the commit message preview