Introducing Policy Cards

Machine-Readable Runtime Governance for Autonomous AI Agents

Policy Cards are a new governance artifact designed for the deployment layer of autonomous AI systems. They attach directly to an agent and specify what it must, may, and must not do at runtime.

Building on the ideas behind Data Cards and Model Cards, Policy Cards define:

• operational constraints and obligations
• evidence and audit requirements
• safety and compliance rules
• cross-links to frameworks (NIST AI RMF, ISO/IEC 42001, EU AI Act)

This initial release includes the full JSON Schema, validator, multiple exemplars (finance, healthcare, defence), and a standards crosswalk.

arXiv preprint: https://arxiv.org/abs/2510.24383

GitHub repository: https://github.com/Symbiotic-Dynamics/policy-cards