A convention that almost did not happen

The platform economy has grown into a workforce that the World Bank estimates at between 154 and 435 million people. Yet for most, the introduction of an app into the work arrangement has functioned as a near-automatic exclusion from labor protection. The model has ‘succeeded’ on the basis of the systematic misclassification of workers as independent contractors, the means by which platforms shed their obligations to respect the minimum-wage, contribute to social security, and respect fundamental rights such as non-discrimination and the rights to organize and bargain collectively. At the same time, algorithmic systems exercise management control over pay rates and task allocation, and decisions related to suspension or deactivation, with none of the accountability of a traditional employer.

In recent years, courts around the world have stepped in and, in most cases, have rectified misclassification. Governments have experimented with regulation to cover at least some workers in the platform economy, with some recognizing workers as employees, others extending some rights and benefits but retaining the self-employed classification and others creating a third category. Into this landscape, the ILO dedicated a tripartite expert meeting on platform work in October 2022.  Unusually, the meeting ended without conclusions amid strong employer opposition and an eventual late-night walkout.

In 2023, a decision was taken by the ILO Governing Body to negotiate a new standard on digital platforms, again over strong employer objection. The first negotiations in 2025 produced little beyond definitions and scope. Only at the International Labor Conference in June 2026 did the Convention cross the line. On June 12, the Conference adopted the Decent Work in the Platform Economy Convention (No. 193) — the first binding international labor standard for the platform economy. For those whose work is organized through an app, this is a major achievement.

Protection that does not depend on classification

The Convention’s strength is that it is status-agnostic. A “digital platform worker” is defined without regard to classification of status in employment (Article 1(b)). The instrument applies to all platforms and all platform workers (Article 2(1)), and the core protections attach to the worker qua platform worker. The Convention escapes the trap that has defeated so much national law, namely the conditioning of all protection on a contested and easily manipulated employment label. Whether a worker is also an employee becomes relevant only to the content of one status-contingent guarantee, namely pay and cost recovery (Article 10), never to the threshold question of coverage. The major compromise contained in the Convention is that the guarantee of remuneration (which requires that workers earnings, excluding any tips or other gratuities, are not lower than the minimum wage) and compensation for expenses or other costs, applies only to employees. States are however directed to consider the extension of these rights to the legitimately self-employed.

Crucially, intermediaries also fall within the scope of the Convention with States having a clear obligation to determine and allocate their respective responsibilities to ensure compliance with the Convention.

The primacy of the facts

Article 9 is one of the core components of the convention. It requires States to ensure the correct classification of platform workers “guided mainly by the facts relating to the performance of work.” This is, in slightly different but functionally equivalent words, the adoption of the primacy-of-facts principle long articulated in the Employment Relationship Recommendation, 2006 (No. 198). As such, the starting point is the actual working relationship, not the labels in a platform’s standard-form contract. This aligns the Convention with the law and practice in many national jurisdictions. The equally authoritative Spanish and French texts use a word translating to “primarily” rather than “mainly,” confirming that Article 9 states a genuine primacy of the facts test.

The enumerated factors are non-exhaustive (“among other elements”), permitting national authorities to weigh the control, integration and economic dependence characteristic of algorithmic management. While Article 9 does not require a rebuttable presumption of employment, the duty to “ensure” correct classification strongly supports one. A State facing widespread misclassification would struggle to demonstrate compliance by any means short of a presumption or an equally effective mechanism shifting the practical burden onto the platform. In any case, labor inspectorates must be empowered. Spain has already enacted such a presumption for delivery riders and the European Union has adopted one of general application.

A first for algorithmic management

The other pioneering development sits in Articles 13 to 16 — the first provisions of any binding international labor standard to regulate algorithmic management directly. Together they establish an international floor for algorithmic accountability at work.

• Transparency (Article 13) requires platforms to inform workers and their representatives about automated systems used to monitor, evaluate or make decisions about work. The express inclusion of representatives gives algorithmic accountability a collective dimension, opening the operation of algorithms to union scrutiny and bargaining.
  
  
• Responsible use (Article 14) ties automated systems to the fundamental principles and rights at work. This means that algorithms may not be deployed to detect and penalize union activity, to discriminate, or to drive work intensities incompatible with a safe and healthy workplace.
  
  
• Explanation and human review (Article 15) grants workers a right, on request, to a written explanation of significant adverse decisions and to a review of decisions resulting in non-payment, suspension, deactivation or termination — with platforms required to ensure “appropriate human involvement.” That phrase can only mean genuine, substantive review by a person competent and authorized to alter the outcome, not a token or rubber-stamping exercise. A nominal “human in the loop” will not satisfy it.
  
  
• Data protection (Article 16) establishes safeguards for workers’ personal data, purpose limitation, and rights of access, rectification and erasure. Read protectively, it constrains what platforms may collect and how they may use it — barring, for instance, the use of location data to frustrate organising, or behavioural data to drive algorithmic wage discrimination.
  
  
• Suspension or Deactivation (Article 17) ensuresthat workers’ accounts cannot be suspended or deactivated on ‘discriminatory or otherwise unlawful grounds’, the latter incorporating the general principles of law of good faith, impartiality and equity

What Next?

The instrument is the subject of negotiation and compromise and does not resolve all problems in the platform economy.  Issues such as the presumption of employment and extending the remuneration floor (Article 10(2)) to economically dependent but self-employed workers, will be left to be fought in national implementation. Its protective potential will hinge on a number of factors.  First, states must ratify the convention. Only the United States and New Zealand voted against the adoption of the convention, which reflects an otherwise broad support among governments. After two states ratify the convention, it will enter into force. Second, legislatures and courts will need to adopt a purposive construction of the text which is both generous as to coverage and narrow as to its potential limitations. Third, the convention will require supervision by the ILO’s Committee of Experts on the Application of Conventions and Recommendations to develop a body of jurisprudence to fill in the gaps in this principles-based instrument. And, most importantly, it will require workers and their unions to use the convention to build collective power.