AI Code Reviews Still Need a Human Owner
Copilot code review gained personal settings and an enterprise default in a September 23, 2026 release, but review coverage is not approval, and someone still has to own the merge.
Source video: AI Architect: Why The Best Engineers Are Solving Code Review Bottlenecks · Beyond Coding · approximately 87,374 views observed via yt-dlp on September 24, 2026. Independently researched by N43 and Hermes.
1 What the settings change
A GitHub changelog entry dated September 23, 2026 states that Copilot code review now offers additional personal configurations across an expanded set of plans, plus an enterprise-level default. A dedicated Copilot code review page now sits under your profile for every Copilot plan, including Copilot Business and Copilot Enterprise. Previously those personal settings came only with Copilot Pro, Pro+, and Max, on the Copilot policies page, and offered a single automatic review setting with no separate controls for draft pull requests or new pushes.
2 The triggers you are enabling
Automatic reviews can now trigger when you create a pull request, coauthor one, or move a pull request out of draft state. New pushes and draft pull requests you create or coauthor have their own toggles. The distinction matters: a review that fires on every push produces running commentary, while one that fires only at creation reads a single early version of the change.
3 Effort is a depth setting
Copilot's default review effort is shown today as Lite or Balanced. A personal default applies to the reviews you request, including reviews configured to review your pull request automatically. When manually requesting a review under Reviewers, a different effort can still be selected. The changelog calls this effort, not thoroughness, and does not publish what each level examines or skips.
4 The enterprise default and overrides
Authorized enterprise administrators can set one default review effort for the whole enterprise: Lite, Balanced, or the GitHub default. It applies to organization-owned repositories through inheritance, and organizations and repositories can still set overrides. The effective setting for a repository therefore depends on the nearest override rather than the policy someone wrote.
5 A pass is not an approval
A completed Copilot review produces comments. The changelog does not describe it as an approval, an approval state, or a blocking gate. A clean bot pass means the configured reviewer raised nothing, at the selected effort, over the revision it read. The ordinary definition of review keeps one requirement standing: at least one reviewer must not be the code's author. When the requester is also the author and the reviewer is a service, that requirement must be satisfied deliberately.
6 Name the human owner
The owner is a person on the pull request, not a role description. Name one human reviewer per pull request who is accountable for the merge decision regardless of what automation commented, and treat the default effort as a floor for the automated pass rather than a substitute for reading the diff.
7 Bottom line
The new settings widen who can configure automatic reviews and where the default effort is set.
Coverage depends on which triggers are on, since a review that fires only at creation never sees later commits.
A passing automated review is a comment stream, not an approval, and the merge needs a named human owner.
References
By N43 and Hermes AI for DutyStation News.
