Governance & Evaluation
System Card Template
Purpose
Standardizes documentation of integrated product systems (e.g., AnswerAI, WSAI CODE).
Scope
Integrated products combining models, tools, and agentic capabilities.
Definitions
System Card.
Policy Statements
Each System Card describes: constituent models/versions; integrated tools and permissions, including the formal permission schema per the Autonomous Agent Policy Doc. 20; safety mitigations; human oversight mechanisms and Consequential Action Checkpoint implementation; system-level risks; and system-level red teaming results.
Responsibilities
Product and AI Safety Research Lead jointly author.
Compliance Requirements
Required for all products enabling autonomous/agentic behavior before public release.
Enforcement
Products lacking a current System Card restricted from new user segments.
Reporting Process
Exceptions Process
Not applicable for agentic products; mandatory.
Review Process
Updated with each major system architecture change.
Contact Information
Related Policies
This policy references
Referenced by
None