Instrument comparison

Compare instruments by what they do, not by the principles they mention.

Similar words can hide different authority, scope, and consequences. The framework gives unlike instruments a common map: which object each acts on, at what level, with what binding force, and what it leaves unresolved.

10 instruments · prototype set

Instrument	Force	Family	Acts on	Level	Editorial caution
NIST AI RMF 1.0 NIST · Public technical framework	voluntary	Risk frameworks	Safeguards Institutions	Corporate	A framework for practice, not a substitute for legal duties or rights analysis. It acts on safeguards and internal institutions; it does not change external incentives by itself.
OECD AI Principles OECD · Intergovernmental recommendation	recommendatory	International institutions	Institutions Incentives	International	Broad policy orientation whose meaning depends on domestic implementation. Its incentive effect runs through adherent governments, not directly on firms.
UNESCO AI Ethics Recommendation UNESCO · Intergovernmental recommendation	recommendatory	Rights frameworks	Institutions Safeguards	International	Its breadth is a feature; implementation still requires institutional translation into named safeguards.
EU Artificial Intelligence Act European Union · Regulation	binding	Laws	Safeguards Institutions Incentives Infrastructures	Regional	Obligations and application dates vary by provision; always verify the current legal text and implementation timeline. It acts on incentives through liability exposure and market access, and on infrastructure by using a training-compute threshold as the trigger for systemic-risk obligations on general-purpose model providers (source-level reading of Article 51; passage review pending).
Council of Europe Framework Convention on AI Council of Europe · International treaty (CETS No. 225)	binding	Treaties	Institutions Safeguards	International	Binding only for parties for which it has entered into force, and largely framed as obligations on states to adopt measures. Verify signature, ratification, and entry-into-force status before relying on it; the overview page, not the treaty text, was reviewed.
OMB M-25-22 AI acquisition memo US Office of Management and Budget · Executive branch policy memorandum	binding	Procurement	Incentives Safeguards	National	Binding on federal agencies, not on the market at large. It works through purchasing power: what vendors must accept to sell to government.
EU Product Liability Directive European Union · EU directive, transposed into national law	binding	Liability	Incentives	Regional	Applies from 9 December 2026. It does not set design rules; it changes incentives by assigning the cost of defects.
IndiaAI Mission Government of India · Cabinet-approved spending programme	voluntary	Public investment and infrastructure policy	Infrastructures Incentives	National	A funding programme rather than a rule: participation is voluntary. It acts on who can build AI by changing access to infrastructure.
Canada Directive on Automated Decision-Making Treasury Board of Canada Secretariat · Treasury Board policy instrument	binding	Audits and evaluations	Safeguards Institutions	National	Binds federal departments only. Also in the GovSpec gold set, so its requirements can be traced clause by clause there.
ISO/IEC 42001 AI management system ISO and IEC · International consensus standard	voluntary	Standards	Institutions Safeguards	International	Voluntary unless a contract or regulator requires it. Certification shows a process exists, not that a given system is safe.

Instrument families and the objects they act on

From the framework: what each family of instrument contributes, and which object it primarily acts on. A family missing from the register above is a gap in this prototype, not in the field.

Risk frameworks1 in register
Help identify and manage harms through repeatable organizational practice.
SafeguardsInstitutions
Laws1 in register
Establish enforceable duties and boundaries, and the authorities that supervise them.
SafeguardsInstitutionsIncentives
Treaties1 in register
Bind parties to shared obligations across borders once ratified and in force.
InstitutionsSafeguards
Standards1 in register
Convert expectations into repeatable technical or organizational requirements.
Safeguards
Audits and evaluations1 in register
Produce evidence, and depend on the capacity to test.
SafeguardsInfrastructures
Procurement1 in register
Changes incentives and market access through the terms of public purchasing.
IncentivesInstitutions
Liability1 in register
Assigns consequences, so it changes what actors document, test, and release.
Incentives
Public investment and infrastructure policy1 in register
Shape capability and access: compute, data, energy, talent, and public digital infrastructure.
InfrastructuresIncentives
International institutions1 in register
Coordinate across borders, share assessment, and address representation.
InstitutionsIncentives
Rights frameworks1 in register
Constrain what governance may legitimately permit.
InstitutionsSafeguards

Organization and AI-system lifecycle

NIST AI RMF 1.0

Risk management functions and outcomes · Trustworthiness and organizational risk practice

Govern
Map
Measure
Manage

International policy and AI actors

OECD AI Principles

Values-based principles and policy recommendations · Trustworthy AI, democratic values, and interoperable policy

Set shared principles
Guide policymakers
Support interoperability

Global public policy

UNESCO AI Ethics Recommendation

Values, principles, and policy action areas · Human dignity, rights, social and environmental well-being

Set normative foundation
Guide policy action
Support readiness assessment

EU market and regulated uses

EU Artificial Intelligence Act

Risk categories, regulated actors, systems, and models · Market rules, safety, fundamental rights, and innovation

Prohibit certain practices
Set duties
Supervise and enforce

Parties' activities within the lifecycle of AI systems

Council of Europe Framework Convention on AI

Obligations of parties, principles, remedies, and procedural safeguards · Human rights, democracy, and the rule of law

Bind parties to principles across the AI lifecycle
Require documentation and complaint routes for affected persons
Establish follow-up on implementation

Federal executive agency AI contracts

OMB M-25-22 AI acquisition memo

Contract terms and solicitations · Competitive markets, data protection and performance

Steers AI contract terms
Protects government data
Limits vendor lock-in
Requires performance monitoring

Products placed on the EU market, including software

EU Product Liability Directive

Defective products and the damage they cause · Compensation for harm

Defines software as a product
No-fault liability for defects
Compensation for harmed people

National AI ecosystem

IndiaAI Mission

Public funds for compute, data, models and skills · Access and domestic capability

Funds public AI compute
Builds a datasets platform
Funds startups and skills
Backs domestic models

Federal government automated decisions

Canada Directive on Automated Decision-Making

Automated decision systems and their impact level · Administrative fairness and transparency

Mandatory impact assessment
Published assessment results
Expert peer review
Mitigations scaled to impact

Any organization that develops, provides or uses AI

ISO/IEC 42001 AI management system

Organizational management system · Repeatable governance practice and certification

Management system requirements
Third-party certification
Continual improvement