Input and Output Validation for AI Agent Components
Validate inputs to and outputs from AI agent tool calls, plugins, and retrieval pipelines to prevent injection attacks from crossing trust boundaries.
validationagentsinjection preventionMCP
Operational boundaries
Controls Catalogue
Control patterns that turn AI use into observable work: owners, boundaries, approvals, logs, exceptions, and evidence.
Off-Agent Authorization for Tool Calls
A control pattern that enforces identity-bound authorization and policy decisions outside the agent runtime before tool calls execute.
authorizationaccess controlagentsMCPprompt injectiontool invocation
Approved AI Tool Register
A maintained record of approved AI tools, allowed use cases, owners, data limits, account requirements, and review status.
inventorygovernanceshadow AI
Data Classification Before AI Use
Classifying data before AI use defines what may be shared, transformed, summarized, retained, or logged.
data securityclassificationprivacy
Human Approval for High-Impact Actions
Human approval prevents sensitive AI-assisted actions from moving directly from model output to execution.
agentsoversightoperations
Prompt and Output Logging
Prompt and output logging preserves enough interaction evidence to support review, monitoring, audit, and incident response.
loggingmonitoringaudit
SSO and MFA for Enterprise AI Platforms
Enterprise identity controls reduce account takeover risk and improve visibility for approved AI tools.
identityMFAenterprise AI