GitHub releases AI security scans for code changes, but findings cannot block merges
AI Scan checks languages and frameworks outside CodeQL’s coverage. It runs inside pull requests, but excludes fork submissions and Dependabot-created changes.
Loading page…
AI Scan checks languages and frameworks outside CodeQL’s coverage. It runs inside pull requests, but excludes fork submissions and Dependabot-created changes.
Listen to this story
Teams can add GitHub’s AI Scan to pull-request review for code in languages and frameworks beyond CodeQL’s current coverage, with results shown alongside CodeQL and sometimes paired with suggested fixes. The October 15, 2026 general-availability release keeps findings advisory: they do not fail code-scanning checks or satisfy merge-blocking rules. Scans require AI Scan to be enabled separately, skip fork and Dependabot pull requests, and use Code Security AI Credits. Private and internal organization repositories also need Code Security or Advanced Security, but no Copilot license is required.
Supported examples include PHP, Shell/Bash, Terraform in HCL, Dockerfiles, JSP, and Blazor.
Automatic scans run only when both code scanning and the repository’s effective AI Scan setting are enabled.
Findings appear in the pull request’s Conversation and Files changed tabs, marked with an AI indicator.
GitHub is broadening automated security review without giving its AI findings the power to stop code changes. On October 15, 2026, it made AI Scan for pull requests generally available. GitHub says the tool finds vulnerabilities beyond CodeQL’s current language and framework coverage, but its findings remain advisory rather than mandatory barriers to merging code.
AI Scan complements CodeQL rather than replacing it. It examines changes in pull requests—proposals to merge code into a repository—and targets languages and frameworks that CodeQL does not currently cover. GitHub lists PHP, Shell/Bash, Terraform configuration written in HCL, Dockerfiles, JSP for Java and Blazor for C# among the supported examples.
The scan runs automatically when both code scanning and the repository’s effective AI Scan setting are enabled. Enabling code scanning alone is not enough: AI Scan must also be enabled under the settings that apply to the repository.
Developers see the findings alongside CodeQL results in the pull request’s Conversation and Files changed tabs. An AI indicator distinguishes them from CodeQL alerts. Findings can include an explanation of the security risk and a suggested autofix.
Automatic scanning does not mean every pull request is eligible. AI Scan does not run on pull requests from forks or on pull requests created by Dependabot. Those exclusions concern whether a scan happens at all, rather than what developers can do with a finding after it appears.
Availability depends on the repository’s visibility and ownership. GitHub lists two access paths, with a security-product requirement for private and internal repositories owned by organizations:
A GitHub Copilot license is not required. That does not remove usage-based billing: AI Scan consumes Code Security AI Credits. Access without Copilot should not be confused with access without a credit cost.
GitHub places that consumption in the AI usage section of enterprise billing and licensing, under the Advanced Security product. Administrators can also set budgets for the specific Code Security AI Credits billing category.
Loading discussion...
Join the conversation
Explain when a warning should require a fix rather than a review.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.