What happened
At GitHub Universe (Oct 28, 2026), GitHub introduced Code scanning for AI applications in public preview: it auto-runs on pull requests and flags indirect prompt injection (XPIA), sensitive information disclosure, and excessive agency in agent code, with remediation guidance inline in the PR. It requires GitHub Advanced Security and is not yet on GitHub Enterprise Server.
Why it matters
This makes AI-specific risk detection (OWASP LLM Top 10) native to the dominant code-hosting workflow at PR time, rather than a bolt-on scanner, and pairs with GitHub's companion previews (AI Scan for repositories, AI-detected secrets in push protection, package firewall).
Applicability
Any team using GitHub Advanced Security and shipping agents/LLM apps should enable it now to gate indirect prompt injection before merge. Evaluate in pull-request review pipelines; hold for GHES-only shops.