DutyGraphAI governance directory

Agent security & threat detection

Google Cloud Model Armor

Google Cloud service that screens LLM prompts and responses for prompt injection, jailbreaks, unsafe content and sensitive data, optionally returning sanitised text. Integrations extend screening to Google-managed MCP server traffic and the Gemini Enterprise agent platform, while the Agent Gateway integration is documented as preview.

commercial · generally available · Research snapshot 2026-09-06

Visit the official product source ↗

Where it fits

Agent security & threat detection · Runtime authorization & controls · Data governance & privacy

Useful conversation with: Cloud security architect, AI platform owner, CISO.

Ask for a demonstration

Show me Model Armor floor settings screening traffic to a Google-managed MCP server, blocking an injected prompt, and clarify which agent integrations are GA versus preview.

Capabilities and evidence

Support labels reflect the supplied research. Documentation and vendor claims are not independent product tests. “Not established” means the researcher did not find support; it does not prove a capability is absent.

Documented by provider

Model Armor inspects incoming prompts and generated responses, can return sanitised versions, and blocks content when prompt injection or jailbreak detection is triggered.

Limit: The overview does not document tool-call authorization or agent action control.

Source s1

Documented by provider

Release notes state integration with Google and Google Cloud MCP servers and with the Gemini Enterprise Agent Platform is generally available, floor settings define baseline filters for MCP server traffic, and Agent Gateway integration is in preview.

Limit: Preview features may change; coverage of third-party MCP servers is not established.

Source s2

Documented by provider

Prompt injection and jailbreak detection flags threats including system instruction manipulation, unauthorized action execution and sensitive information retrieval.

Limit: Detection thresholds are configurable and effectiveness is not quantified.

Source s2

Limitations to discuss

Sources

  1. Model Armor overview · Google Cloud · official docs
    Access date reported by researcher: 2026-09-06
  2. Model Armor release notes · Google Cloud · official release
    Access date reported by researcher: 2026-09-06

Listing does not imply partnership, supplier status, a working DutyGraph integration, or a compliance certification.

Suggest a correction