Check validation limits before assigning security alerts to GitHub agentic autofix

Who does this general availability announcement cover?

The release announcement covers Code Security or Advanced Security customers on GitHub Enterprise Cloud, including data residency environments. This article discusses adoption within that announced scope.

The concept documentation describes broader eligibility and the audience for the existing Autofix experience. The two sources describe different scopes. Do not assume every repository type mentioned in the documentation is covered on the same terms by this Enterprise Cloud announcement.

How does it differ from existing Copilot Autofix?

Existing Copilot Autofix suggests a code change for a supported alert for you to review and apply. Agentic autofix explores relevant files, makes changes, attempts validation, and opens a draft PR. Enabling it replaces Generate Fix with Assign to Copilot on individual alerts.

Only default-branch alerts are supported. Alerts found only on pull requests or other branches cannot be assigned. You can also select up to 25 alerts from an alert list or security campaign for a single PR. There is no requirement to fill that maximum.

What does rerunning CodeQL verify?

The agent reruns CodeQL with the code scanning query suite to check its changes. This is best effort. The documentation says it cannot confirm fixes for custom CodeQL queries, the security-extended query suite, or third-party tool alerts. An open PR alone does not establish that those alerts are resolved.

For review, consider checking the PR's validation steps alongside its changed files. Use team tests to check whether the input that reproduced the issue is handled correctly and whether related features still work. Alert resolution and preserved functionality are separate checks.

Which policies and costs should you check first?

Under the announcement's requirements, an enterprise administrator must allow Agentic autofix in Advanced Security policies before organization or repository administrators enable it in Code Security settings. A Copilot billing method must also be defined. Do not assume a repository setting can override a higher-level restriction.

Usage consumes AI Credits from the Code Security budget and GitHub Actions minutes. Existing Copilot Autofix suggestions do not consume AI Credits, so the two have different cost conditions. For a team trial, consider a small group of related alerts and name both a reviewer and someone responsible for checking costs.

After a run, record the query suite used, the rerun result, and alerts that remain unverified in the PR's validation record. Before the next assignment, agree on who will check those remaining alerts and with which checks to reduce review stalls. Official sources were checked on October 17, 2026.