GitHub's agentic vulnerability scanner no longer stops at the lines a pull request changes. Pondero reports that, per GitHub's changelog dated 28 October 2026, AI Scan for repositories can now crawl an entire codebase and surface issues in languages CodeQL has never supported — a scope jump from the pull-request-only version GitHub shipped in September.

The repository-wide version entered public preview on github.com and GitHub Enterprise Cloud with data residency, according to the same report. Once an enterprise enables the AI Scan policy, organization owners and security managers turn it on for their org; personal-account owners enable it in repository settings. A scan analyzes the latest commit on the default branch, with the agentic scanner exploring the repository and focusing on the code most likely to hold vulnerabilities, running within a fixed time budget to keep cost bounded.

Findings land in the normal code scanning alert list alongside CodeQL alerts, support assignees and linked pull requests, and can be fixed with agentic autofix. Where AI Scan and CodeQL cover the same repository, AI Scan skips what CodeQL already reports, and no GitHub Copilot license is required. Public repositories get the feature without a Copilot seat; private and internal repositories still need GitHub Code Security or GitHub Advanced Security, and personal accounts cannot run it on private repos at all.

Two caveats are worth noting before teams lean on it. GitHub has not published false-positive rates or example findings for the repository-wide scan, and rescans do not auto-close fixed alerts — a security team has to kick off a fresh scan after merging a fix to clear its backlog. Still, the shift matters: the old scanner guarded what was changing, while this one audits what already exists.

Sources