Independent · Evidence-based · Born at Oxford

Certification for the human labour behind AI.

There is nothing artificial about the millions of people who annotate data and moderate content to power the systems we call intelligent. Fairwork independently certifies the firms that commission this work, based on evidence including the voices of the workers themselves. For those firms, it's independent proof of responsible AI that stands up to regulators, investors and customers.

A non-profit spun out of the Oxford Internet Institute (OII) & Berlin Social Science Center (WZB).
F A I R P A Y · F A I R C O N D I T I O N S · F A I R C O N T R A C T S · F A I R M A N A G E M E N T · F A I R R E P R E S E N T A T I O N ·
16M+
Workers whose conditions improved
880+
Company ratings
440+
Verified improvements delivered
41
Countries where we have verified conditions

Seven years of Fairwork research. Now a certification programme built for AI supply chains.

The problem

The workforce behind AI is hidden, and increasingly regulated.

AI relies on outsourced data work in regions where labour protections are thin, monitoring is hard, and responsibility becomes diffuse across layers of suppliers.

Laws like the EU's Corporate Sustainability Due Diligence Directive (CS3D) now extend accountability across supply chains, including digital services and AI. Ethical oversight is becoming a legal obligation.

What certification delivers
Regulation

Be regulation-ready

Align early with global standards and avoid compliance risks.

Reputation

Earn trust

Show stakeholders your AI practices match your values.

Quality

Stronger teams

Fair wages and secure jobs help retain skilled data workers.

Growth

Financial upside

Tap into ESG capital, new markets, and long-term savings.

See why firms certify
What we certify

A VERIFIED ACCOUNT OF HOW WORK IS ACTUALLY DONE.

Governance

The lead firm

We assess how a firm structures, resources, procures and oversees the AI work it commissions: the decisions that shape working conditions upstream.

Outcomes

The suppliers

We independently verify the conditions under which suppliers deliver the work.

Voice

Stakeholder Inclusion

Anonymous, non-retaliatory worker testimony sits at the centre of every assessment. No certification decision rests on a single source of evidence.

Five principles

The standard every assessment is measured against.

01

Fair Pay

Workers earn at least the local living wage after work-related costs.

02

Fair Conditions

Safe working environments, paid leave, and basic protections against occupational risk.

03

Fair Contracts

Secure contracts, clear terms, and genuine employment stability.

04

Fair Management

Fair systems for data oversight, feedback, appeals and redress.

05

Fair Representation

Support for workers' voice, organising, and good-faith negotiation.

Two levels of certification

From independent assessment to verified improvement.

Entry level · Transitional

Fairwork Onboard

Valid 12 months
  • The project has undergone independent verification
  • Supplier evidence and worker testimony collected
  • Findings accepted within the certified scope
  • A binding, time-bound Corrective Action Plan in place
Higher level · Verified

Fairwork Endorsed

Valid as long as progress is made
  • Agreed corrective actions implemented
  • Identified harms remedied and independently verified
  • Continuous, worker-level improvements supported by evidence
  • Oversight and procurement systems working in practice

Onboard confirms that independent scrutiny has taken place and a remediation plan is in motion. Endorsed confirms that improvements are verified and continuing.

Recognition

The Fairwork principles are endorsed by governments, businesses and institutions around the world.

German Federal Government UK FCDO University of Oxford WZB Berlin GIZ Good Business Charter Internet Freedom Foundation Lagos Business School IIIT Bangalore McDonald's Germany The Church of England

Endorsements shown are illustrative of Fairwork's existing recognition, see full list here.

Let's build fairer AI, together.

Book a free consultation
Why certify

If you rely on AI, you rely on a supply chain that should be fair.

Certification is for more than AI companies. If your business trains models, operates platforms, or delivers AI-powered services, human data work sits somewhere in your supply chain. Certifying it helps you prepare for due-diligence law, earns the trust of investors and customers, and can build a more stable workforce.

01 · Regulation-ready

Meet due-diligence law early

Align with CS3D and global due-diligence regimes before compliance becomes urgent or negligence gets fined. A confidential assessment supports internal due diligence and regulatory readiness.

02 · Trust

Prove your practices match your values

Show investors, partners and customers that your AI is sourced responsibly, backed by independent evidence and worker testimony.

03 · Resilience

Retain skilled data workers

Fair pay and secure work produce a more stable, skilled and reliable data workforce, which is critical to long-term AI quality.

04 · Financial upside

Move beyond checkbox compliance

Access ESG-aligned capital and new markets, and build supplier relationships designed for continuous improvement rather than one-off audits.

Who it's for

Lead firms with the leverage to shape how work is done.

Certification is awarded to the lead firm that commissions AI work through suppliers. You're eligible where you:

Commission AI-related labour through one or more suppliers.

Can identify the suppliers materially involved in delivering the work.

Have contractual or commercial leverage to require supplier cooperation.

Agree to open the project to independent verification, including worker outreach.

Certification can't proceed where supplier access is denied, worker outreach is obstructed, or the proposed scope is too narrow to support a credible assessment.

Scope

What certification covers.

Certification applies to a defined project or body of contracted work, not a firm's entire global operations, unless expressly agreed.

In scope

  • Data annotation & enrichment
  • Content moderation
  • Validation & quality assurance
  • Transcription
  • AI-enabled customer support
  • Contact-centre operations supporting automated systems

Typically out of scope

  • Core infrastructure & compute
  • Model development & research roles
  • Specialised technical engineering

These involve different employment structures, risk profiles and access constraints, and fall outside the standard scope.

Not sure if your work is in scope?

Talk to our team
How certification works

Rigorous by design. Built to improve working conditions.

Every certification is conducted against the Fairwork AI Principles using a triangulated methodology. No decision rests on any single source of evidence.

Methodology

Three streams of evidence, weighed together.

Stream 1

Desk research

Legal, regulatory, sectoral and wage-benchmark context for every supplier location, shaping the evidence requests and sampling intensity.

Stream 2

Management evidence

Contracts, payroll, grievance records, procurement and oversight systems, plus interviews with lead-firm and supplier management.

Stream 3

Worker evidence

Anonymous, voluntary, safeguarded worker outreach. Non-retaliation is a mandatory condition of engagement.

The two levels

From Onboard to Endorsed.

Entry · Transitional · 12 months

Fairwork Onboard

Independent scrutiny + a binding remediation pathway
  • Project independently examined
  • Worker testimony incorporated
  • Findings accepted; accountability established
  • Corrective Action Plan agreed where required
Higher · Verified · Ongoing

Fairwork Endorsed

Continuous improvement proven in practice
  • Corrective actions implemented
  • Harms remedied to meet Fairwork standards
  • Worker improvements evidenced
  • Oversight systems demonstrably operational
Two ways to engage

Start with a snapshot, or commit to the full pathway.

Pathway A

Audit-only

A one-time, confidential assessment of labour standards in your AI supply chain. You receive a detailed evaluation against the Fairwork Principles, benchmarking, and a tailored action plan. A powerful first step towards fairness and due-diligence readiness.

Pathway B

Full certification

Commit to the complete process. Receive Fairwork Onboard after independent assessment, then earn Fairwork Endorsed once improvements are verified: a public, evidence-backed signal of ethical leadership.

The assessment journey

From application to certified, with worker outreach at its core.

Stage 1–2

Scoping & context

Confirm eligibility, define the certified scope, identify suppliers, and set the legal and wage-benchmark context.

≈ 4–8 weeks
Stage 3

Management evidence

Review governance, contracts, payroll and oversight systems; interview lead-firm and supplier management.

≈ 6–10 weeks
Stage 4

Worker verification

Safeguarded, anonymous worker outreach to verify conditions in practice, protected by non-retaliation.

≈ 6–10 weeks
Stage 5–6

Findings & CAP

Triangulate evidence, share draft findings for factual review, and agree a time-bound Corrective Action Plan.

≈ 7–12 weeks
Stage 7–9

Decision & follow-up

Share the outcome, verify remediation, and monitor continuously toward Endorsed status.

ongoing
What the mark means

A certification you can verify.

Every use of the certification mark states its level, its scope and its validity period. Certification is time-limited and applies only to the certified scope.

Where publication is agreed, Fairwork publishes a report covering the certified scope, the methodology, key findings against each principle, and the outcome. The mark then links back to that report.

Ready to map your AI labour footprint?

Book a consultation
Impact & evidence

The evidence behind the standard.

Fairwork certification is built on years of academic research into digital labour: a proven, independent methodology that gives your certification real weight with regulators, investors and customers.

16M+
Workers whose conditions improved
880+
Company ratings
440+
Verified improvements delivered
41
Countries where we have verified conditions

Figures illustrative of Fairwork's cumulative research impact. See the full list here.

From research to certification

Why the programme exists.

For years, Fairwork benchmarked labour conditions publicly against its principles. Benchmarking increases transparency, but on its own it doesn't embed standards into procurement or create sustained accountability.

Certification closes that gap: it combines independent verification, worker testimony and structured, time-bound remediation, requiring firms to open real projects to scrutiny and act on what's found.

Case studies

What engaging with Fairwork delivers.

Sama
Kenya · Uganda · Data annotation

The Workers Behind AI at Sama

The first Fairwork AI report used the ratings framework to assess conditions at a data-annotation company with a social-impact mission, showing that fairness at work isn't guaranteed, but that meaningful, evidenced improvements are achievable through engagement.

HITL
Bulgaria · W. Asia · sub-Saharan Africa

Humans in the Loop: Who Powers AI?

A complex, multi-tier subcontracting model assessed against the Fairwork AI Principles. The company's willingness to connect the research team with subcontractors and workers made verified improvements possible, even in conflict-affected settings.

Recognition

The Fairwork principles are endorsed by governments, businesses and institutions around the world.

German Federal Government UK FCDO University of Oxford WZB Berlin GIZ Good Business Charter Internet Freedom Foundation Lagos Business School IIIT Bangalore McDonald's Germany The Church of England

Endorsements shown are illustrative of Fairwork's existing recognition, see full list here.

See where your supply chain stands.

Request an assessment
Team

The people behind the standard.

Fairwork Certification is led by the team that built the Fairwork programme. Between them, they have assessed working conditions at hundreds of companies in more than 40 countries and negotiated hundreds of verified improvements for workers.

Mark Graham
Executive Director

Mark Graham

Oxford Internet Institute, University of Oxford

Mark founded Fairwork and has led its growth into a programme that has assessed companies in more than 40 countries and negotiated hundreds of improvements to working conditions. He is Professor of Internet Geography at the Oxford Internet Institute, University of Oxford. His books include Feeding the Machine, on the hidden human labour behind AI. As Executive Director, he leads the strategy and standards of the certification body.

Patrick Feuerstein
CFO and Director

Patrick Feuerstein

Berlin Social Science Center (WZB)

Patrick co-leads Fairwork Germany and directs the Fairwork Observatory, working directly with companies, workers and policymakers to turn assessment findings into agreed improvements. He is a sociologist at the Berlin Social Science Center (WZB) and completed his doctorate at the University of Göttingen. As CFO, he oversees the finances and operations of the certification body.

Richard Heeks
Director

Richard Heeks

Global Development Institute, University of Manchester

Richard has spent three decades advising governments and international organisations, including the FCDO, GIZ and the ILO, on how digital technology changes work and development. He co-developed the Fairwork framework and led its extension across the Global South. He is Professor of Digital Development at the University of Manchester and Director of its Centre for Digital Development.

Janaki Srinivasan
Member

Janaki Srinivasan

University of Oxford

Janaki has spent years assessing platform companies and advocating for better conditions for their workers as a co-investigator of Fairwork India, one of the programme's longest-running country teams. She is Associate Professor of Digital South Asian Studies at the University of Oxford, and previously taught at IIIT Bangalore, where she convened its Centre for IT and Public Policy. Her book The Political Lives of Information was published by MIT Press.

Talk to the team.

Book a consultation
About

An independent certifier, by design.

Fairwork began as a research initiative of the Oxford Internet Institute and the WZB Berlin Social Science Center, evaluating working conditions across the platform and AI economy against five principles of fair work.

The certification programme now operates as an independent non-profit, a company limited by guarantee, carrying that research heritage into a formal assurance model for AI supply chains. The structure is deliberate: with no shareholders to answer to, our only incentive is the integrity of the assessment.

IndependentA non-profit CLG with no commercial stake in the outcome of any certification.
Evidence-ledDecisions rest on triangulated evidence, never on a firm's self-declaration alone.
Worker-centredAnonymous worker testimony and non-retaliation are non-negotiable conditions of engagement.
TransparentWhere publication is agreed, findings are published and every mark links to its report.
AccountableCertification is time-limited, subject to monitoring, and can be suspended or revoked.
Get in touch

Let's build fairer AI, together.

Whether you're beginning to map your AI labour footprint or advancing toward certification, we'll help you find a clear, credible path. Book a free consultation and we'll talk through scope, timelines and pathway.

info@fair.work
linkedin.com/company/fairwork
fair.work