AI Runtime Security Tools

Google renames AI Protection findings tied to Model Armor

Google Cloud's Security Command Center release notes record that on 24 Sep 2026 AI Protection finding categories were renamed to drop the legacy VERTEX_1P_ prefix and make clear they cover tuned models, including the finding for tuned models not protected by Model Armor. Teams that filter on category names need to update.

Facts read on .

On 24 Sep 2026, Google Cloud's Security Command Center release notes listed a change to AI Protection finding names. The legacy VERTEX_1P_ prefix was removed, so VERTEX_1P_TUNED_MODEL_DETECTED became TUNED_MODEL_DETECTED, and VERTEX_1P_TUNED_MODEL_NOT_PROTECTED_BY_MODEL_ARMOR became TUNED_MODEL_NOT_PROTECTED_BY_MODEL_ARMOR. Google says the new names clarify that AI Protection detects tuned models.

The second finding matters for runtime security because it reports a tuned model that is not covered by Model Armor, Google's built-in screening for prompts and responses. Alert rules, dashboards and SIEM filters keyed to the old category strings will not match the new ones.

The same release notes show earlier 2026 changes to AI Protection: general availability at the organization level in the Premium tier on 5 Mar 2026, project-level activation on 27 Jul 2026 and data residency in Saudi Arabia on 13 Aug 2026.

Sources