AI security is moving from scanning to active validation
A GitHub and evidence cohort tests whether agentic security tools reproduce vulnerabilities, verify exploits, and control remediation rather than only flagging suspicious code.
Supported by the targeted cohort
18 subject-filtered implementations show that active validation is real but still early. Exploit or vulnerability reproduction appears in 27.8% of the cohort, measured evaluation in 33.3%, and controlled remediation in 0%. The evidence supports a shift beyond scanning, but not yet a mature end-to-end security-operations category.
Market snapshot
Comparable measurements from independent market surfaces.
Implementations and packages
Deduplicated and subject-filtered primary cohort.
Median repository stars
Calculated across repositories with at least one star.
With active exploit validation
5 items in the subject-filtered cohort.
With measured evaluation
6 items in the subject-filtered cohort.
Search-match dynamics
Bars show monthly GitHub total_count plus captured Hugging Face models; the right column shows downloaded items used for detailed analysis.
What exists inside the category
One item may contain more than one feature.
Scanning and detection
6 · 33.3%
Active exploit validation
5 · 27.8%
Controlled remediation
0 · 0%
Measured security evaluation
6 · 33.3%
CI integration
5 · 27.8%
Human security approval
1 · 5.6%
Reproducible evidence reports
6 · 33.3%
Representative projects
Cross-source validation
These publications are not part of the primary numeric cohort.