ILO Paper: AI Ethics Codes Discuss Work but Rarely Cite Labour Standards

More than 60% of the AI ethics guidelines reviewed by ILO researchers discuss work, yet fewer than 20% mention the International Labour Organization or its standards. [Fact] That gap is the starting point of ILO Working Paper 182, "AI Ethics and the ILO's international labour standards system: exploring synergies," published on October 23, 2026. [Fact] The authors are Jordi Agustí-Panareda, a socio-legal researcher and the ILO's Ethics Officer, and Daniel Samaan, a senior economist in the ILO Research and Statistics Department. [Fact]

Five principles against a century of labour law

The paper takes the five principles of the Floridi and Cowls framework from 2019, widely used in AI ethics: beneficence, non-maleficence, autonomy, justice, and explicability. [Fact] It maps them against the ILO Constitution, the Declaration of Philadelphia, and the Declaration on Fundamental Principles and Rights at Work. [Fact] The method is an interpretive comparison, with no survey or dataset. [Fact]

The guideline figures come from a companion review by Samaan and colleagues, still in press, which covers 245 documents. [Fact] Applied to all 245, the two shares mean roughly 147 guidelines that talk about work and fewer than 49 that cite the ILO. [Estimate] In that review, the world of work does not appear as a topic of its own among the seven thematic clusters found. [Fact]

Where the two agree, and where they split

The authors find strong overlap on justice, beneficence, and non-maleficence: fair distribution, well-being, and protection of workers' life and health map onto existing ILO principles. [Fact]

The split comes on two points. AI ethics frames autonomy as an individual's power to decide, while the ILO system adds a collective version through freedom of association and collective bargaining. [Fact] For the right to collective bargaining, the authors "do not see a clear ethical equivalent" in existing AI ethics frameworks. [Fact]

The second point runs the other way. The ILO "has historically placed less emphasis on privacy as a labour right," the paper says, and its code on data privacy dates from 1997, written before current AI systems. [Fact] Data collection, algorithmic management, workplace surveillance, and opaque hiring decisions are the areas where the authors expect AI ethics to push the application of labour standards forward. [Claim]

A hiring case as the test

The paper describes a pending lawsuit in an unnamed ILO member state. [Fact] The complaint alleges that a recruiting platform built applicant profiles from résumés, employer data, third-party online sources, and inferred personal traits, then supplied "match scores" and rankings to employers without letting applicants see, correct, or contest the data. [Fact] The challenge was filed under consumer-reporting law, not under labour law written for AI hiring. [Fact] The authors point to the Discrimination (Employment and Occupation) Convention, 1958 (No. 111) as the benchmark that fits the dispute. [Claim]

The paper also notes that the ILO adopted the Decent Work in the Platform Economy Convention, 2026 (No. 193). [Fact]

What to prepare

For human resources specialists and human resources managers, the hiring case lists the records a screening tool should be able to produce: which data sources were used, which traits were inferred, how the score was built, and why a candidate was rejected. [Claim] Teams that can show those four items for every automated rejection are ready for the kind of claim the paper describes. [Claim]

For labor relations specialists, the paper's central finding concerns bargaining: AI ethics codes leave out the collective channel, so the introduction of monitoring or scoring tools is a subject unions and employers can put on the table themselves. [Claim]

For compliance officers, the gap between a 1997 data-privacy code and current workplace AI is the area to track as the ILO considers guidance or review of standards. [Claim]

Limits

This is a working paper of exploratory, interpretive analysis by two ILO staff, and the views are the authors', not the ILO's. It relies on one AI ethics framework, which the authors say could yield a different mapping if swapped, and the guideline statistics come from a review that is not yet published; the authors also disclose that ChatGPT was used for drafting assistance.

Sources

• ILO News, "How can international labour standards help shape ethical AI at work?", October 23, 2026: https://www.ilo.org/resource/news/how-can-international-labour-standards-help-shape-ethical-ai-work
• Agustí-Panareda, J., Samaan, D., "AI Ethics and the ILO's international labour standards system: exploring synergies," ILO Working Paper 182, Geneva: ILO, 2026, https://doi.org/10.54394/00035324 (CC BY 4.0, © ILO).

About this analysis

This article was written with AI assistance from the full text of ILO Working Paper 182. Short quotes are from the paper; the summary is ours and has not been reviewed or endorsed by the ILO. The guideline counts marked [Estimate] are our arithmetic on the paper's percentages.