Top 10 Principles for Ethical Artificial Intelligence

Top 10 Principles for Ethical Artificial Intelligence is the 11 February 2017 document through which UNI Global Union, the global trade union federation representing 20 million workers across 150 countries in skills and services sectors — ICT, finance, commerce, media, gaming, postal and logistics, cleaning and security, and others — entered the AI ethics field as one of the earliest organised-labour bodies to do so. The document operationalises UNI's core demand that "artificial intelligence must put people and planet first," providing unions, shop stewards, and workers with concrete, actionable demands on AI transparency and application in the workplace and in society. Published through UNI's The Future World of Work platform, it was later listed in the OECD AI Policy Observatory's catalogue of ethical AI tools and referenced by UNI in its May 2019 response to the first OECD AI Recommendation as the worker-side framework that pre-dated and informed intergovernmental AI principles work.

UNI Global Union: mandate and scope

UNI Global Union was formed on January 1, 2000 through the merger of four global union federations: the International Federation of Commercial, Clerical, Professional and Technical Employees (FIET), Media and Entertainment International (MEI), the International Graphical Federation (IGF), and Communications International (CI). It operated under the name "Union Network International" until March 2, 2009, when it adopted its current name. Headquartered in Nyon, Switzerland, UNI affiliates unions in 150 countries and represents workers across the services and skills sectors most directly exposed to AI-driven displacement and management: ICT and telecommunications, finance and banking, retail commerce, media and entertainment, postal services, gaming, and cleaning and security, among others. Its sectoral reach gives its AI principles a distinctly different register from national technology-worker unions: UNI speaks for the service worker who experiences AI primarily as algorithmic management, automated customer interaction, workplace surveillance, and job displacement — not as a technical development process they participate in shaping.

The ten principles

The document sets out ten demands UNI frames as the minimum requirements for ethical AI in the workplace and in society:

1. Transparency. Workers must have the right to demand transparency in AI systems' decisions, outcomes, and underlying algorithms. AI deployment must include worker consultation from the design stage, not only at deployment.

2. Ethical Black Box. AI systems must be required to incorporate a recording mechanism — an "ethical black box" — that captures information about the system's operation to ensure accountability, facilitate audit, and support redress when harms occur.

3. Legal Responsibility. Legal responsibility for an AI system's actions must rest with a human person — its designers, deployers, or operators. Systems must be designed to comply with existing laws, including fundamental rights and privacy protections. Machines may not hold legal standing.

4. Worker Rights Alignment. Every AI system's deployment must be evaluated against its compatibility with workers' fundamental rights; AI expansion may not proceed independently of that check.

5. Protection from Displacement. Governments must provide social security coverage, continuous lifelong learning and retraining pathways, and targeted measures supporting workers displaced by AI — a just-transition obligation rather than a market outcome.

6. Worker Consultation. Workers must be systematically consulted on the implementation, development, and deployment of AI systems that affect their working conditions; consultation is a procedural right rather than an optional step.

7. Multi-stakeholder Governance. UNI calls for the establishment of multi-stakeholder governance bodies at global and regional levels, including AI designers, developers, researchers, manufacturers, employers, lawyers, civil society organisations, and trade unions, with built-in whistleblowing mechanisms and monitoring procedures to enforce compliance.

8. Ban on Lethal Autonomous Weapons. UNI urges a global ban on lethal autonomous weapons systems, including autonomous cyber warfare capabilities.

9. Regulatory Compliance. AI must be developed to be responsible, safe, and useful; machines must retain the legal status of tools under human control and may not be granted independent agency or authority.

10. Global Convention. UNI calls for a binding global convention on ethical AI — an intergovernmental treaty framework — as the mechanism through which the preceding demands can be made enforceable across jurisdictions and the unintended negative consequences of AI addressed at scale.

Context: The Future World of Work and the leadership summit

The principles document emerged from UNI's engagement with digital transformation through The Future World of Work strategic platform and from a leadership summit at which trade union leaders from across the world called for a global convention on ethical AI. UNI framed the document's urgency this way: "As Artificial intelligence (AI), robotics, data and machine learning enter our workplaces across the world displacing and disrupting workers and jobs, unions must get involved." Philip Jennings served as UNI General Secretary until 2018, when Christy Hoffman succeeded him; the 2017 principles document was published during Jennings' tenure as the document through which UNI committed to a concrete, negotiable AI agenda for its member unions.

Reception and intergovernmental reach

When the OECD published its first AI Recommendation in May 2019 — the first intergovernmental AI standard adopted by G20 governments — UNI General Secretary Christy Hoffman responded directly: "The application of A.I. should not be self-regulated and left to the big tech companies. The principles announced today create a much needed regulatory framework for national governments." Christina Colclough, UNI's Director of Platform and Agency Workers, Digitalization and Trade, noted that UNI had worked with TUAC (the Trade Union Advisory Committee to the OECD) as part of the process producing the OECD standard. The OECD AI Policy Observatory subsequently listed the UNI principles in its curated inventory of ethical AI governance tools. The UC Berkeley Labor Center's September 2025 analysis A First Look at Labor's AI Values found UNI's AI positions among the most comprehensively developed in the global labor movement, noting UNI's coverage of AI ethics, algorithmic management, and data protection across multiple documents, and treating UNI's transparency principle as foundational to the broader framework of labor's AI governance demands.

Position in the corpus

This publication fills the global labor federation perspective slot in the corpus's worker-organizing movement area — the gap the Synthesizer flagged when queuing it. The corpus already holds the worker-organizing strand through the Managed by Bots report, the Authors Guild open letter, and the Artists 4 Safe AI letter — all sectoral or campaign-level documents. What was missing was a publication representing the global labor federation layer: the organisational structure that links national unions across sectors and countries into a coordinated position on AI as a system-level worker-rights problem. UNI's 2017 principles are that entry — among the earliest sustained attempts by a global union federation to codify AI as requiring union-to-union coordination and intergovernmental treaty-making, pre-dating the OECD AI Principles by two years and most national union AI statements by several years.

The document's register within the corpus is the pre-deployment worker-voice demand: its core insistence is that worker consultation must precede AI deployment, that transparency is a right unions can demand in negotiations rather than a feature companies provide voluntarily, and that AI-driven displacement requires proactive government commitment rather than post-hoc adjustment. This distinguishes it from the algorithmic-management litigation strand — which contests deployed systems through GDPR and employment law — while providing the upstream policy framework within which that litigation finds its normative footing.