A neon software dashboard shows AI processing code modules, flagging issues, and securing the results.
GitHub has made it faster to launch AI Scans across many repositories. After you enable the setting, you can start one-off scans of up to 20 repositories. That cuts the repeated clicking, but it is still manual, on-demand scanning. It is not continuous coverage.

A neon software dashboard shows AI processing code modules, flagging issues, and securing the results. What changed​

GitHub's November 16, 2026 changelog says the bulk start is available in two places:

  • When you enable the organization-level setting. You can choose which repositories to scan at that point.
  • From the code scanning organization alerts page. You can start scans for a batch of repositories there too.

You can still start a scan for a single repository from its own alerts page. The bulk-start capability is in public preview, so it may change. GitHub lists availability as github.com and GitHub Enterprise Cloud with data residency. It does not name GitHub Enterprise Server. A Copilot license is not required.

What AI Scan for repositories does​

AI Scan looks for security vulnerabilities on a repository's default branch. GitHub says it covers languages CodeQL doesn't support. It also finds additional issues in languages CodeQL already analyzes. Findings land in the same code scanning backlog as CodeQL alerts. That means the usual workflows apply: assigning alerts, linking issues, security campaigns and agentic autofix.

This is separate from AI Scan for pull requests, which has its own setting. Enabling one does not enable the other.

Enabling the setting does not start a scan​

GitHub's documentation is explicit about this. The setting only makes scanning possible. Someone with write access still has to start the scans. The bulk start only changes how many repositories you can launch at once. GitHub gives no figures for time saved, and it doesn't say whether results for a batch arrive together.

Limits to know before you launch a batch​

GitHub's documentation lists several limits:

  • Default branch only. Other branches are not analyzed.
  • Manual triggers only. Scheduled scans and scans on push are not supported. Self-hosted runners are not supported either.
  • Not exhaustive. The agent picks which files and functions to examine within a time budget. GitHub says a scan doesn't guarantee it finds every vulnerability.
  • Possible false positives and duplicates. AI-generated findings can be inaccurate and need human review. CodeQL alerts are used to suppress matching findings, but that matching is best effort, so duplicates can occur.
  • No automatic re-scan after a fix. During the preview, merging a fix doesn't trigger another scan. An alert closes automatically only if the fix changes the exact line that triggered it. Otherwise you close it by hand.

Eligibility, policy and cost​

The cost side is where batch launches can bite. Each scan consumes Code Security AI credits. GitHub's documentation says detection steps are capped at 30 minutes and 150,000 AI credits. It also says if a hard budget leaves no Code Security AI credits available, AI Scan cannot produce findings. "No Copilot license required" is not the same as free.

Repository eligibility, from GitHub's documentation and its October 28 launch notice:

OwnerVisibilityAvailability
OrganizationPublicAvailable
OrganizationPrivate or internalNeeds GitHub Code Security or GitHub Advanced Security
Personal accountPublicAvailable
Personal accountPrivateNot available

Policy controls sit above the repositories. Enterprise owners can allow or prevent organizations from enabling AI Scan, and allowing it does not enable it automatically. The documentation says the policy defaults to "Not allowed." Organization owners and security managers can turn the setting on. Repository admins can opt individual organization-owned repositories out.

A sensible rollout​

  1. Confirm the enterprise policy allows AI Scan.
  2. Check which of your repositories are eligible. Private and internal ones need Code Security or Advanced Security.
  3. Review your included AI credits and set budgets for metered usage. GitHub recommends doing this before enabling the feature.
  4. Start with a batch of high-value repositories, especially ones in languages CodeQL doesn't cover. This is a practical suggestion, not GitHub guidance.
  5. Triage the findings and filter by the AI Scan tool. Treat each one as unverified until a person has reviewed it.
  6. After merging fixes, start a new scan to confirm they worked and to close alerts.

Related automation​

Bulk start is not the only recent addition. GitHub also added a REST API control. You can now turn AI Scan for repositories on or off with the REST API at the organization and repository levels. Scripts can handle the enablement setting at scale. The 20-repository batch limit applies to starting scans from the UI.

Analysis​

This is an incremental but useful change for security teams with large estates. A one-by-one start process is a real barrier when you want to try a new scanner across dozens of repositories. A batch of 20 reduces that friction.

The caveats matter just as much. The scan is manual, covers one branch, has a time budget and uses metered credits. It works best as a supplement to CodeQL, secret scanning and normal review. It does not replace them. Public preview status also means behavior and limits may change.

 

References

  1. Bulk start AI Scan for repositories GitHub Changelog 2026-11-16T15:50:22+00:00
  2. AI Scan for a repository's backlog - GitHub Docs docs.github.com
  3. AI Scan for repositories in public preview - GitHub Changelog github.blog