Codiatech Blog
How to run a controlled GitHub AI Scan pilot across repositories
GitHub has exposed repository AI Scan settings through its REST API. Plan where to pilot it, who can trigger scans, and how your team will review findings.

On November 13, 2026, GitHub announced REST API controls for repository AI Scan at the organization and repository levels. Teams can now inspect and change these settings in bulk. The feature is in public preview. For a security team planning an initial rollout, the practical questions are where to enable it and who will run the first scans. GitHub changelog
What the new API setting controls
Use GET to read the setting and PATCH to change it. The organization endpoint is /orgs/{org}/code-scanning/ai-scan. The repository endpoint is /repos/{owner}/{repo}/code-scanning/ai-scan. The new repository_scan field governs repository backlog scanning, while pr_scan governs pull request scanning. Once repository scanning is enabled, someone with write access can start an on-demand scan of the default branch. GitHub REST API reference
Check which repositories inherit the policy
When an organization disables repository AI Scan, individual repositories cannot opt in. Enabling it makes eligible repositories inherit the setting and allows repository admins to opt out. Before piloting with a handful of repositories, review the broader group that will inherit the enabled setting. Record who authorizes the organization policy and who is responsible for actually running scans.
Start by capturing the stored settings with GET. A practical first group might be one API service and one background worker, each with a named reviewer. If you need to change the organization setting, agree on the resulting organization-wide eligibility or start in an organization where that setting is already appropriate. An enterprise policy that prevents enablement produces a 422. If either requested change is blocked in a combined request, both settings stay as they were. Inspect the error and send it to the policy owner. GitHub’s policy and error behavior
Turn the first scan into a review process
Ask a designated contributor with write access to trigger the first scan. Each run consumes Code Security AI credits, so identify the relevant budget and who will review usage before running it. For each finding, capture the affected code path, the conditions required to reach it, its potential impact, and a way to confirm the behavior. Review overlapping findings alongside existing CodeQL alerts.

After merging a fix, run the relevant regression tests and check the finding’s status. A merge does not start a new AI Scan during the preview. Plan a follow-up run when you need one. The current preview excludes scheduled scans, automatic scans on push, and self-hosted runners. GitHub’s repository AI Scan documentation
Decide whether to expand the pilot
Review four measures with the team. Count the findings reviewed and confirmed, then compare review time and credits consumed. Record who took ownership of findings in each repository. Expand in small groups when the team can act on the findings within its review capacity. Keep your existing code security checks in place. GitHub says the scan selects portions of the default branch to analyze within a time budget.
Organization-owned public repositories are eligible. Private or internal repositories require GitHub Code Security or GitHub Advanced Security, and the announced management feature does not support GitHub Enterprise Server. Confirm your repository eligibility, plan and organization policy before setting a rollout date. GitHub’s eligibility requirements
Codiatech’s cloud and DevOps services can help you assess where repository policy and security findings fit into your deployment workflows.
Primary source: github.blog