Skip to main content
Bureau of AI
Category: Trustworthy AI Principles

UNESCO Recommendation on the Ethics of Artificial Intelligence

Also known as: Recommendation on the Ethics of Artificial Intelligence, UNESCO Recommendation on AI Ethics, UNESCO AI Ethics Recommendation
Simply put

The UNESCO Recommendation on the Ethics of Artificial Intelligence is a global standard-setting instrument, described in the evidence as the first of its kind, that sets out shared ethical principles for how AI systems should be developed and used. It places the protection of human rights and human dignity at its center and promotes values consistent with human rights, human dignity, and protection of the environment and ecosystems. Its stated aim is to help ensure AI systems work for the good of humanity, individuals, societies, and the environment.

Formal definition

The UNESCO Recommendation on the Ethics of Artificial Intelligence is a standard-setting instrument issued through UNESCO that articulates fundamental principles—such as transparency—and values grounded in the promotion and protection of human rights, human dignity, and environmental and ecosystem protection. According to the evidence, the protection of human rights and dignity is described as the cornerstone of the instrument, and it is characterized as the first global standard-setting instrument for ethical AI. As a UNESCO recommendation, it functions as a normative, values-based framework rather than a jurisdiction-specific binding law; it should not be conflated with mandatory regulatory regimes or with prescriptive model risk management guidance. The evidence provided does not detail its specific principle taxonomy, implementation mechanisms, adoption date, or enforcement status, so those aspects are out of scope for this entry.

Why it matters

The UNESCO Recommendation on the Ethics of Artificial Intelligence matters because it is characterized in the evidence as the first global standard-setting instrument for ethical AI, giving it a distinctive role in the international landscape. Unlike jurisdiction-specific regulatory regimes, it articulates shared, values-based principles—centered on the protection of human rights and human dignity—that can inform how organizations and governments frame their own AI governance approaches across borders. For professionals building governance programs, it offers a common ethical vocabulary that reaches beyond any single legal system.

Its significance is normative rather than legally binding in the way domestic statutes or supervisory guidance are. A UNESCO recommendation functions as a values-based framework that member states and organizations may draw on and translate into their own instruments; it does not, on its own, create the kind of enforceable obligations found in binding law or in prescriptive model risk management guidance. Governance and compliance teams should therefore treat it as a reference point for ethical alignment rather than as a source of specific, auditable requirements.

Who it's relevant to

AI governance and policy specialists
Those designing organizational AI governance frameworks can reference the Recommendation as an internationally recognized set of ethical principles centered on human rights, human dignity, and environmental protection. It is useful as a values anchor when harmonizing internal policy across multiple jurisdictions, provided it is treated as a normative reference rather than as binding law.
Compliance and legal professionals
Compliance and legal teams should understand that, as a UNESCO recommendation, the instrument functions as a values-based framework and should not be conflated with mandatory regulatory regimes. Where organizations operate in jurisdictions that have drawn on its principles, legal professionals may need to trace how those principles were translated into domestic instruments, which is beyond the scope of the evidence here.
Model risk managers and validators
The Recommendation is not a prescriptive model risk management guidance document and does not substitute for validation, verification, or control frameworks. Model risk professionals may find its transparency and human-rights principles helpful in framing the ethical context of model use, but it does not provide the measurable, auditable requirements that model risk programs typically rely on.
Multinational organizations and public-sector bodies
Organizations and governments operating across borders may use the Recommendation as a shared ethical baseline given its characterization as the first global standard-setting instrument for ethical AI. Its explicit goal of making AI systems work for the good of humanity, individuals, societies, and the environment makes it a useful common reference where jurisdiction-specific rules diverge.

Inside UNESCO Recommendation on the Ethics of Artificial Intelligence

Non-binding recommendation instrument
The UNESCO Recommendation on the Ethics of Artificial Intelligence is adopted by UNESCO Member States as a recommendation rather than a treaty or enforceable law. It sets out shared values and principles for AI ethics that member states are encouraged to reflect in national policy, but it does not by itself create binding legal obligations.
Values and principles framework
The instrument articulates high-level ethical values (such as human dignity, human rights, and environmental well-being) and principles intended to guide the design, development, and deployment of AI systems. These are stated as normative aspirations rather than as prescriptive technical controls.
Policy action areas
It identifies areas where member states are encouraged to translate ethical principles into concrete policy, which may span governance, data policy, and oversight. Practitioners should treat these as directional guidance for national implementation rather than as a compliance checklist.
Global and cross-sector scope
The Recommendation is intended to apply broadly across countries and sectors that adopt it, distinguishing it from sector-specific model risk guidance or jurisdiction-specific law. Its actual effect in any given jurisdiction depends on how, and whether, individual member states implement it.
Ethics-oriented rather than model-risk-oriented
Its focus is on the ethics and societal governance of AI. It is not a model risk management standard in the sense of guidance historically framed for measuring, monitoring, and controlling model risk, and it does not prescribe validation or monitoring methodologies.

Common questions

Answers to the questions practitioners most commonly ask about UNESCO Recommendation on the Ethics of Artificial Intelligence.

Is the UNESCO Recommendation on the Ethics of AI legally binding on the countries that adopted it?
No. As commonly characterized, the Recommendation is a non-binding normative instrument rather than a treaty or enforceable law. Its adoption reflects a political and ethical commitment by member states, but it does not by itself create directly enforceable legal obligations. Any binding effect would arise only if a jurisdiction chose to implement its principles through its own national legislation or regulation, and the specifics of such implementation vary by country. Readers should not treat the Recommendation as equivalent to binding frameworks.
Is the UNESCO Recommendation interchangeable with instruments like the NIST AI Risk Management Framework, ISO/IEC 42001, or the EU AI Act?
No, and treating them as interchangeable is a common error. These instruments differ in issuing body, legal status, and purpose. The UNESCO Recommendation is a values-and-principles instrument adopted through UNESCO. It is distinct in nature from a voluntary technical standard, a voluntary risk-management framework, or a binding regional law. They may address overlapping themes such as transparency or accountability, but they are not substitutes for one another and operate at different levels of specificity and enforceability. Organizations should map how each instrument applies to their context rather than assuming one satisfies the others.
How does an organization begin translating the Recommendation's high-level principles into internal AI governance practices?
Because the Recommendation is principle-based rather than prescriptive, organizations typically treat it as a source of values to inform, not replace, their governance structures. A common approach is to map its principles to existing internal policies, roles, and oversight bodies, then identify gaps. This work sits within AI governance (organizational structures, policies, and accountability for AI systems) and should be distinguished from model risk management activities, which focus on identifying, measuring, monitoring, and controlling risks from specific models. The Recommendation does not specify particular control procedures, so operational detail must come from an organization's own frameworks.
Does aligning with the Recommendation demonstrate regulatory compliance?
Not on its own. Alignment with a non-binding ethical instrument is generally distinct from compliance with applicable binding law in a given jurisdiction. An organization operating under specific national or sectoral regulation would still need to satisfy those legal requirements separately. Where a jurisdiction has incorporated aspects of the Recommendation's principles into its own law or supervisory expectations, alignment may support compliance, but professionals should verify what is actually binding in their jurisdiction rather than assuming the Recommendation carries legal force.
How might the Recommendation relate to model risk management rather than only broad AI ethics?
The Recommendation's principles can inform governance expectations that surround model risk management without themselves prescribing risk-management methodology. For example, principles concerning transparency or accountability may reinforce an organization's rationale for practices such as documentation, oversight, or monitoring. However, the operational discipline of model risk management, including activities framed in some sectors by guidance such as SR 11-7 in U.S. banking, remains a separate and more specific body of practice. The Recommendation should be treated as a values input, not as model risk management guidance.
Who within an organization typically owns responsibility for acting on the Recommendation's principles?
Ownership varies by organization and is not specified by the instrument itself. In many governance arrangements, responsibility is distributed across roles and lines of defense, with policy and oversight functions setting expectations and operational teams implementing them. Because the Recommendation is high-level, organizations generally assign accountability through their existing governance structures rather than expecting the instrument to define roles. Where and how this responsibility is allocated is an internal design decision that depends on organizational size, sector, and regulatory context.

Common misconceptions

The UNESCO Recommendation is binding law that organizations must comply with.
As commonly understood, it is a non-binding recommendation adopted by UNESCO Member States. It can influence national policy and inform organizational practice, but on its own it does not impose enforceable legal obligations; any binding effect arises only through separate national laws that a member state may enact.
It is interchangeable with frameworks like the NIST AI RMF, ISO/IEC 42001, or the EU AI Act.
These instruments differ in issuer, nature, and scope. The UNESCO Recommendation is an ethics-oriented instrument adopted through UNESCO, whereas others are voluntary standards, voluntary frameworks, or binding regional law issued by different bodies. They should not be treated as equivalent or as substitutes for one another.
Adopting the Recommendation constitutes a model risk management program.
The Recommendation addresses AI ethics and governance at a values-and-principles level. It does not supply the identification, measurement, monitoring, and control processes associated with model risk management, so it cannot substitute for a dedicated model risk framework.

Best practices

Treat the Recommendation as directional ethics guidance and confirm the specific legal obligations that actually apply in your jurisdiction through the relevant national laws before relying on it for compliance.
Map the Recommendation's values and principles to your existing governance policies to identify gaps, rather than adopting it as a standalone compliance checklist.
Keep the ethics-oriented aims of the Recommendation distinct from your model risk management controls, and maintain separate processes for validation, monitoring, and control of model risk.
Avoid presenting alignment with the Recommendation as equivalent to conformance with voluntary standards or compliance with binding regional law; document each framework's status and issuer separately.
Use qualified language in internal documentation when describing how the Recommendation influences your controls, noting that its effect depends on national implementation.
Track how member states in your operating jurisdictions implement the Recommendation over time, since its practical impact evolves through subsequent national policy rather than through the instrument itself.