GitHub makes AI pull-request scanning generally available, but findings cannot block merges

GitHub has made AI Scan for pull requests generally available, adding vulnerability checks for code that CodeQL does not currently cover. The catch for teams hoping to make it a gatekeeper: its findings are advisory and cannot block a merge.

Watch Desk analysis

What happened

GitHub says AI Scan can run automatically on pull requests when code scanning and the repository’s effective AI Scan setting are enabled. It covers languages and formats including PHP, Shell/Bash, Terraform configuration, Dockerfiles, JSP for Java and Blazor for C, and more. Findings appear alongside CodeQL results, marked with an “AI” indicator, and can include an explanation of the risk and a suggested fix.

The feature is available for public repositories owned by organisations or personal accounts. Private and internal organisation-owned repositories require GitHub Code Security or GitHub Advanced Security. A Copilot licence is not required, but AI Scan uses Code Security AI Credits. It does not run on fork pull requests or pull requests created by Dependabot.

Why it matters

This gives developers another way to surface potential vulnerabilities in changes, including in areas outside CodeQL’s current coverage. But a finding is a prompt to investigate, not a safety catch that can stop risky code reaching the main branch. Teams will also need to account for AI Credit usage, even without a Copilot subscription.

Our read

Useful extra coverage, with the limits in plain sight. Teams can try AI Scan as another reviewer in the pull-request workflow, but should not mistake an advisory alert for an enforced security check. The distinction between “found something” and “stopped the merge” is rather more than a footnote.

What to watch

• Which additional languages and frameworks GitHub brings into AI Scan.
• How teams manage Code Security AI Credit use as scanning scales.
• Whether GitHub adds ways to make AI Scan findings enforceable in merge rules.

Discussion spark: Should AI-generated security findings ever be allowed to block a pull-request merge, or should that decision always require a human or a separate, stricter check?

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.