Governance & Evaluation
Safety Report Template
Purpose
Standardizes internal and external safety reporting for model releases.
Scope
All Level 2-4 model releases under the AI Safety Framework.
Definitions
Safety Report.
Policy Statements
Each report documents: risk classification rationale (citing Doc. 24's table); evaluation results across all four Evaluation Framework domains; red team findings and remediation status; CBRN-gate outcome where triggered; residual risk acceptance decisions with signatories and expiration dates; post-deployment monitoring plan.
Responsibilities
AI Safety Research Lead owns compilation; Governance Review Board reviews before approval.
Compliance Requirements
Required prior to any Level 2+ release gate approval.
Enforcement
Releases without a completed report are blocked.
Reporting Process
Exceptions Process
Not applicable; mandatory for qualifying releases.
Review Process
Compiled per release; retained 5 years.
Contact Information
Related Policies
This policy references
Referenced by
None