Summary
On October 7, 2026, GitHub introduced a model built specifically to detect leaked secrets, aiming to catch exposed credentials and tokens more accurately within its secret scanning.
What changed
GitHub added a dedicated, purpose-built model for detecting leaked secrets in secret scanning.
Why it matters
Specialized detection models can raise recall and cut false positives versus pattern-based scanning, strengthening the credential-leak defenses that matter more as AI agents generate and handle more code and secrets.
Evidence excerpt
GitHub has added a model designed specifically to detect leaked secrets.