What's new?

• Added the gcpModelArmorFloorSetting prevention type for tracking the GCP Model Armor Floor Setting enforced on each project — the minimum prompt and response filtering that Model Armor templates cannot undercut. The prevention discovery reads the per-project Floor Setting from the GCP > Model Armor > Floor Setting CMDB resource (CMDB-first, with no extra Model Armor API calls) and synthesizes one prevention per project. The Floor Setting is a per-project singleton with no delete API, so the prevention is score-not-delete: only the Guardrails prevention row is removed when the floor is no longer present. Requires @turbot/gcp-modelarmor 5.0.0 or later in the workspace.
• The gcpModelArmorFloorSetting prevention publishes a per-filter view of the live Floor Setting so each filter can be governed on its own: every Detection and Enforce filter as an enforced or not-enforced row, each Responsible AI filter with its minimum confidence level (Low and above, Medium and above, or High), Sensitive Data Protection (Basic), and the floor's integrated services (AI Platform and Google-Managed MCP Servers). Responsible AI filters are scored on a confidence gradient — a stricter confidence level scores higher — rather than a single on/off check, so an objective can require a minimum confidence rather than mere presence.
• Added example files for the 10 per-filter GCP Model Armor Floor Setting objectives. Each example wires the GCP > Model Armor > Floor Setting > Settings control to enforce the floor and sets the corresponding Floor Setting > Settings > Detection, Responsible AI, or Enforce leaf policy that the objective scores against, so a customer can apply an individual filter directly. The Responsible AI examples (Hate Speech, Dangerous Content, Sexually Explicit, Harassment) use Medium and above as the minimum confidence.
• Added a per-filter set of 10 GCP Model Armor Floor Setting objectives so customers can opt into individual filters rather than a single bundle. The new objectives are enforceFloorSettingMaliciousUrlDetectionForGcp, enforceFloorSettingPromptInjectionDetectionForGcp, enforceFloorSettingSensitiveDataProtectionForGcp, enforceFloorSettingHateSpeechFilterForGcp, enforceFloorSettingDangerousContentFilterForGcp, enforceFloorSettingSexuallyExplicitFilterForGcp, enforceFloorSettingHarassmentFilterForGcp, enforceFloorSettingAgentPlatformForGcp, enforceFloorSettingGoogleManagedMcpServersForGcp, and enforceFloorSettingTemplateCreateAndUpdateForGcp. Each objective selects its filter through a key argument under staticMappings.arguments on the gcpModelArmorFloorSetting prevention type, and is scored independently against the live Floor Setting.

Control Types

Added

• GCP > Model Armor > Floor Setting > Prevention
• GCP > Model Armor > Floor Setting > Prevention > Discovery

Prevention Types

Added

• GCP Model Armor Floor Setting

Prevention Objectives

Added

• Enforce Floor Setting Agent Platform (Vertex AI) for GCP
• Enforce Floor Setting Dangerous Content Responsible AI filter for GCP
• Enforce Floor Setting Google Managed MCP Servers integration for GCP
• Enforce Floor Setting Harassment Responsible AI filter for GCP
• Enforce Floor Setting Hate Speech Responsible AI filter for GCP
• Enforce Floor Setting Malicious URL detection for GCP
• Enforce Floor Setting Prompt Injection detection for GCP
• Enforce Floor Setting Sensitive Data Protection for GCP
• Enforce Floor Setting Sexually Explicit Responsible AI filter for GCP
• Enforce Floor Setting Template Create and Update for GCP

Prevention Examples

Added

• Enforce Agent Platform (Vertex AI) integration on GCP Model Armor Floor Setting
• Enforce Dangerous Content Responsible AI filter on GCP Model Armor Floor Setting
• Enforce Google Managed MCP Servers integration on GCP Model Armor Floor Setting
• Enforce Harassment Responsible AI filter on GCP Model Armor Floor Setting
• Enforce Hate Speech Responsible AI filter on GCP Model Armor Floor Setting
• Enforce Malicious URL detection on GCP Model Armor Floor Setting
• Enforce Prompt Injection and Jailbreak detection on GCP Model Armor Floor Setting
• Enforce Sensitive Data Protection (Basic) on GCP Model Armor Floor Setting
• Enforce Sexually Explicit Responsible AI filter on GCP Model Armor Floor Setting
• Enforce Template Create and Update floor on GCP Model Armor Floor Setting