Definition
A library of principles is the codified set of enterprise values, policies, and operating rules that are translated into machine-readable governance instructions for AI agents. Rather than each AI deployment defining its own rules, the library holds the enterprise's authoritative governance standards in one place, from which every agent and workflow draws its behavioral constraints
A library of principles is the codified set of enterprise values, policies, and operating rules that are translated into machine-readable governance instructions for AI agents. It is the organized, structured source of the governance layer: rather than each AI deployment defining its own rules, the library of principles holds the enterprise's authoritative governance standards in one place, from which every agent and workflow draws its behavioral constraints.
Think of the library of principles as the enterprise's constitution for AI behavior. Just as a legal constitution sets the foundational rules from which all other laws derive their authority, the library of principles sets the foundational rules from which all agent policies are derived. An individual agent's specific permissions and constraints are expressions of the broader principles in the library, not independently invented rules.
What does a library of principles contain?
A library of principles typically organizes governance across three levels.
Foundational values: the enterprise's core ethical and operational commitments that govern all AI behavior regardless of use case. These might include commitments to data privacy, transparency, human oversight, and non-discrimination. They apply universally and cannot be overridden by individual agent configurations.
Functional policies: governance rules specific to particular domains or functions. Finance agents may operate under stricter approval requirements than internal communications agents. Legal agents may have different data access permissions than operations agents. Functional policies translate foundational values into domain-specific constraints.
Operational rules: the specific instructions that govern agent behavior in individual workflows: what data can be accessed, what actions can be taken autonomously, what requires escalation, and what must be logged. Operational rules are the most granular layer and the most likely to change as use cases evolve.
Why does a library of principles matter for enterprise AI governance?
Without a library of principles, governance is inevitably ad hoc: each agent deployment makes its own decisions about what rules to enforce, producing inconsistency across the portfolio. A library of principles creates a single source of governance truth that every agent references, making governance coherent, auditable, and updatable from one place.
It also future-proofs governance. As new regulations emerge, new agents are deployed, or organizational values evolve, updates to the library propagate across every agent that references it, without requiring each to be individually reconfigured.
Is a library of principles the same as a corporate code of conduct?
Related but operationally different. A corporate code of conduct is written for human interpretation and compliance. A library of principles translates the same intent into machine-readable instructions that AI agents can follow automatically. The values are often the same; the encoding is different.
Who is responsible for maintaining the library of principles?
Typically a combination of legal, compliance, risk, and AI governance functions, with input from senior leadership on the foundational values. The library is an organizational governance artifact, not a technical one, even though its outputs are technical.
How does the library of principles relate to governance as code?
The library of principles is the organized source of governance intent. Governance as code is the practice of implementing that intent as machine-readable instructions. The library is what you maintain; governance as code is how you deploy it.
