[Development] AI-generated contributions to Qt

[Edward Welbourne]

Axel Spoerl (20 July 2026 10:24) observed:
> Apart from copyright concerns, I find it particularly hard to review
> monster-commits of AI created content without developer ownership.

Note that "monster-commits" are already covered, to some degree at
least, by the "commits should be atomic" language in our [[Commit
Policy]] - which our [[Review Policy]] says reviewers should verify is
being followed.

The [[Commit Policy]] also has

  7. Do not commit anything you do not understand.

which largely covers some of the concerns about AI-generated changes: if
the contributor doesn't understand it then reviewers should reject it.
And the part about atomic commits is within

  9. Produce commits which facilitate reviews and contribute to a useful
     history which can be read, searched, annotated and cherry-picked.

which, again, is an argument for breaking up monster commits into
readily-reviewed pieces.  The [[Review Policy]] also says

  3. Never approve changes outside your expertise.

which implies never approving what you don't understand.  A reviewer
can always assert that as valid grounds to decline to approve.

(Note: [[...]] references are to pages on wiki.qt.io.)

	Eddy.