The ILO’s AI ethics map is a bridge to labour institutions, not a new rule
Working Paper 182 links five common ethics principles to the ILO normative tradition and exposes gaps around privacy, autonomy and collective representation. Organisations can use the map to find accountable instruments without presenting it as binding law.
What happened
ILO Working Paper 182 compared five widely cited AI ethics principles with the ILO’s constitutional values, declarations and fundamental rights at work.
Why it matters
Generic ethics statements often lack owners and remedies; labour institutions can supply decision rights and collective processes, but an exploratory paper does not itself create legal obligations.
Five principles meet a labour framework
ILO Working Paper 182 compares justice, beneficence, non-maleficence, autonomy and explicability with values and rights expressed in the ILO Constitution, the Declaration of Philadelphia and the Declaration on Fundamental Principles and Rights at Work.
The authors find substantial convergence around justice, beneficence and non-maleficence. They also identify different emphases around privacy, autonomy and collective worker and employer representation. That matters because a generic fairness or transparency principle does not specify who can contest a workplace decision, which representative body must be consulted or what remedy follows.
The paper is an exploratory interpretive mapping. Its publication page states that the views are the authors’ and do not necessarily reflect the ILO. It is therefore neither a new international labour standard nor a binding interpretation of an existing one.
Institutions add procedures
An official EESC opinion provides an independent European comparison. It emphasizes information and consultation, human review and social dialogue when AI affects work. An EESC opinion is advisory; it does not create a universal rule, and obligations vary by jurisdiction, sector, system and existing agreements.
Together, the sources suggest a useful governance move. Ethics principles can identify the value at risk, while labour law, collective agreements and industrial-relations processes may identify the decision maker, affected group, evidence duty and channel for challenge. Neither source eliminates the need for legal and worker-representation review.
Build a four-column crosswalk
For every workplace-AI use case, list the ethical principle, the applicable instrument or agreement, the accountable role and the contest or remedy path. Add evidence: system purpose, data and proxy risks, affected tasks, performance by relevant group, review authority and change history. Mark uncertain legal interpretations for counsel rather than converting them into policy facts.
Involve worker representatives early enough to change the design, not only after procurement. Separate individual notice from collective consultation. Record where autonomy is constrained by scheduling, monitoring or performance management, and whether a human reviewer can reverse the decision.
This complements the UK agentic authority ledger: the ledger records what an agent may do; the labour crosswalk records whose rights and institutions constrain that authority.
Decision rule
Use Working Paper 182 as a discovery map. Do not label it a binding standard. A deployment should proceed only after each principle has an applicable owner, evidence, consultation path and jurisdiction-specific legal status.