Feedback issue for the Business Logic Security Analyzer beta
About the Business Logic Security Analyzer
The Business Logic Security Analyzer (BLSA) is an AI agent that finds vulnerabilities in how your application enforces its own rules: broken object-level authorization (BOLA/IDOR, CWE-639), missing authorization (CWE-862), mass assignment (CWE-915) and related access control flaws. Pattern-based SAST often misses these because the code is valid and only the intent is wrong. BLSA scans the whole repository on the default branch, or the files a merge request changes, and reports findings as a SAST report: in the Vulnerability Report, the merge request security widget, and merge request approval policies.
Availability: GitLab Ultimate · GitLab.com (open beta, self-serve); GitLab Self-Managed and GitLab Dedicated in 19.6 · Beta · Each scan consumes GitLab Credits
Docs: https://docs.gitlab.com/user/application_security/business_logic_scanning/
How to turn it on
On GitLab.com, a top-level group Owner turns on GitLab Duo Agent Platform, Experiment and beta features, and the Business Logic Security Scan foundational flow, then applies the Business logic profile to a project from Secure > Security configuration. Merge request scans run automatically on that project; a whole-repository scan starts from GitLab Duo Chat (/flow: and select Business Logic Security Scan) or the API. The model is set per top-level group in GitLab Duo model selection; Claude Sonnet 5.5 is the default.
Share your feedback
Comment on this issue with:
- The codebase scanned (language, framework, rough size) and whether it was a whole-repository or merge request scan
- Which model ran, and the effort tier if you changed it
- Whether findings were real, actionable, or false positives, and how you decided
- Findings you expected and did not get
- What you did with a finding (fixed, dismissed, created an issue, blocked a merge)
- How long the scan took and what it cost in GitLab Credits, and whether that felt proportionate
- Anything that surprised you in enabling it, reading results, or the Vulnerability Report
If you hit an error or a scan that never finished, include the agent session link so Engineering can look it up.
Internal: product requirements and status are in &23435.