Connectors · Google Private Preview

Google Gemini

Rencore monitors Google Gemini across 21 governance policies, 7 reports, and 13 inventories, detecting model access risks, cost overruns, and agent lifecycle issues automatically.

AI & Agents

Published 16 May 2026 For Head of IT, CISO, CIO / CXO

Google Gemini is in private preview. Join the waiting list and we will reach out when access opens up.

Join the waiting list

Rencore Gemini governance is a set of 21 policies, 7 reports, 7 segments, and 13 inventories that audit Google's Vertex AI Platform and Agent Builder for security gaps, cost overruns, and operational risks. It detects models deployed without proper access controls, agents with excessive data permissions, and projects exceeding budget thresholds, giving IT visibility into enterprise Google AI usage.

59 governance capabilities: 13 inventories · 21 policies · 7 reports · 7 segments · 4 automations

Why govern Google Gemini with Rencore

Control model access and permissions

Detect models deployed without proper access controls, projects with overly broad IAM roles, and agents connected to sensitive data sources. Each finding includes severity and recommended remediation.
Track AI spending

Monitor costs across projects, models, and agent invocations. Policies alert when spending exceeds thresholds at the project or organization level. Reports break down costs by model type and team.
Manage agent lifecycle

Identify agents not updated in 90+ days, stale deployments consuming resources, and projects without assigned owners. Reports show agent activity trends and usage patterns.

What Rencore discovers

Rencore automatically inventories these Google Gemini object types.

Gemini Project

Google Cloud project containing Vertex AI and Agent Builder resources
Gemini Model

Custom or tuned ML models registered in Vertex AI Model Registry
Gemini Endpoint

Model serving endpoints that host deployed models for prediction
Gemini Pipeline Job

ML pipeline execution jobs in Vertex AI Pipelines
Gemini Tuning Job

Fine-tuning jobs for foundation models in Vertex AI
Gemini Dataset

Training datasets used for model training and fine-tuning in Vertex AI
Gemini Agent

Deployed AI agents (Reasoning Engines) in Vertex AI Agent Engine
Gemini Notebook Runtime

Managed notebook runtime instances for interactive ML development
Gemini Engine

Search, recommendation, or chat engines in Vertex AI Agent Builder
Gemini Data Store

Data stores for search, RAG, and recommendation in Agent Builder
Gemini Document

Indexed documents within Agent Builder data stores
Gemini Session

Conversation sessions within Agent Builder engines
Gemini User

Users with access to Google AI resources, linked to Microsoft 365 accounts

How Gemini governance works in Rencore

Rencore connects to Google’s Vertex AI Platform and Agent Builder via Google Cloud APIs and inventories projects, models, agents, deployments, and data connections. Policies run on every scan cycle and evaluate each resource against governance rules, flagging security, cost, and lifecycle issues.

The multi-vendor AI governance challenge

Organizations using Google Gemini alongside Microsoft 365 Copilot, OpenAI, and Claude need consistent governance across all AI platforms. Rencore provides a unified governance view, detecting the same categories of risk whether your AI workloads run on Google Cloud, Azure, AWS, or third-party platforms.

Who uses Gemini governance

CISOs use it to enforce access controls on model deployments and data connections. Heads of IT track cost trends and identify optimization opportunities. CIOs use adoption reports to compare Google AI usage with other AI platforms across the organization.

Getting started

Provide Rencore with Google Cloud API credentials scoped to Vertex AI. All 21 policies activate on first scan, covering models, agents, projects, and deployments. No per-project configuration required.

Policies

21 governance rules that detect violations and risks.

Severity Category

Gemini notebook runtime in unhealthy state

Detects notebook runtimes that are reporting an unhealthy health state

High Security
Gemini user is deactivated in Entra ID

Detects Gemini users who are deactivated in the parent Entra ID

Medium Security
Gemini user is external user in Entra ID

Detects Gemini users which are guest in the Entra ID directory

Medium Security
Gemini agent not updated in 90 days

Detects deployed reasoning engines (agents) that have not been updated in the last 90 days

Medium Security
Gemini tuning job in failed state

Detects fine-tuning jobs that have failed

Medium Operation
Gemini pipeline job in failed state

Detects ML pipeline jobs that have failed

Medium Operation
Gemini model not updated in 90 days

Detects models in the registry that have not been modified in the last 90 days

Medium Operation
Gemini endpoint not updated in 90 days

Detects model serving endpoints that have not been modified in the last 90 days

Medium Operation
Gemini engine not updated in 90 days

Detects Agent Builder engines that have not been modified in the last 90 days

Medium Operation
Gemini endpoint with no deployed models

Detects model serving endpoints that have zero models deployed

Medium Sprawl
Gemini notebook runtime idle for 30 days

Detects notebook runtimes that have not been updated in the last 30 days

Medium Costs
Gemini notebook runtime with expensive accelerator

Detects notebook runtimes using high-cost GPU accelerators like A100

Medium Costs
Gemini engine with no connected data stores

Detects Agent Builder engines that have zero data stores connected

Medium Sprawl
Gemini project with too many failed pipeline jobs

Detects projects that have more than 10 failed pipeline jobs

Medium Operation
Gemini user requires offboarding

Detects Gemini users linked to deactivated Entra ID accounts that need offboarding

Medium User Offboarding
Gemini agent without description

Detects deployed reasoning engines (agents) that have no description set

Low Security
Gemini model with no deployments

Detects models that are not deployed to any endpoint

Low Sprawl
Gemini dataset not updated in 180 days

Detects datasets that have not been modified in the last 180 days

Low Sprawl
Gemini dataset without description

Detects datasets that have no description set

Low Sprawl
Gemini endpoint with traffic on single model

Detects endpoints with multiple deployed models but traffic routed to only one

Low Costs
Gemini data store with no documents

Detects data stores that contain zero indexed documents

Low Sprawl

Need a rule that isn't listed? Rencore's Policy Builder lets you create custom policies tailored to your organization. Learn more about the Policy Builder

Reports

7 analytics views and dashboards.

Models per Project

Number of custom models in each Google Cloud project

Bar Chart · Adoption
Endpoints per Project

Number of model endpoints in each Google Cloud project

Bar Chart · Adoption
Datasets per Project

Number of datasets in each Google Cloud project

Bar Chart · Adoption
Tuning Jobs by State

Distribution of Vertex AI model tuning jobs by current state

Donut Chart · Operation
Pipeline Jobs by State

Distribution of Vertex AI pipeline jobs by current state

Donut Chart · Operation
Notebook Runtimes by State

Distribution of Vertex AI notebook runtimes by current runtime state

Donut Chart · Operation
Data Stores by Industry Vertical

Distribution of Agent Builder data stores by industry vertical classification

Bar Chart · Adoption

Automations

4 automated remediation workflows.

Delete Gemini Agent

Automatically deletes a Gemini agent (Reasoning Engine) after approval
Delete Gemini Endpoint

Automatically deletes a Gemini endpoint after approval
Delete Gemini Data Store

Automatically deletes a Gemini data store after approval
Stop Gemini Notebook Runtime

Automatically stops a Gemini notebook runtime after approval

Segments

7 data groupings for targeted filtering.

Active Endpoints

Vertex AI endpoints with at least one deployed model
Failed Pipeline Jobs

Vertex AI pipeline jobs in a failed state
Succeeded Pipeline Jobs

Vertex AI pipeline jobs that completed successfully
Failed Tuning Jobs

Vertex AI model tuning jobs in a failed state
Succeeded Tuning Jobs

Vertex AI model tuning jobs that completed successfully
Active Notebook Runtimes

Vertex AI notebook runtimes currently in a running state
Stopped Notebook Runtimes

Vertex AI notebook runtimes currently in a stopped state

Frequently asked questions

Does Rencore support governance for AI tools beyond Microsoft Copilot?

Yes. Rencore connects to Claude, OpenAI, Gemini, GitHub Copilot, Cursor, Windsurf, AWS Bedrock, Azure AI Foundry, and other AI platforms. Each connector provides tailored policies for cost management, security, adoption tracking, and access control, giving IT a unified governance view across all AI tools the organization uses.

What is Rencore governance?

Rencore governance is a SaaS platform that continuously monitors your Microsoft 365 tenant for policy violations, configuration drift, and security risks across SharePoint, Teams, Power Platform, Copilot, and AI Agents. It automates compliance evidence collection, surfaces oversharing and sprawl, and provides actionable remediation workflows, reducing manual audit effort by up to 80%.

How do Rencore policies work?

Rencore ships with hundreds of pre-built policies that detect governance violations across every connector, oversharing, sprawl, cost overruns, security risks, and compliance gaps. Policies run on a continuous schedule, evaluate each discovered object against configurable rules, and flag violations with severity (High, Medium, Low), category, and a recommended action.

Trusted by