In June 2026, at its 114th session, the International Labour Conference (ILC) adopted Convention 193, setting down international standards for decent work in the platform economy. The Convention was adopted by an overwhelming majority. The governments of the United States and New Zealand voted against its adoption, while fourteen governments abstained, including Australia, Canada, India, Japan, and the United Kingdom. The Convention is the outcome of hard-fought negotiations between workers, employers and governments at the International Labour Organisation (ILO), the only multilateral institution that grants workers’ and employers’ representatives a formal role in international lawmaking. Negotiations over the text began at the previous ILC in 2025, but the process that led to the Convention has a much longer history. It has been shaped by the ILO’s engagement with the platformization of labor since the late 2010s, and more importantly, by over a decade of organizing and advocacy by platform workers, unions and labor movements worldwide.
The crowning achievement of the Convention is its recognition of platform work as work, and the consequent extension of the ILO’s normative framework of Fundamental Principles and Rights at Work to platform-mediated labor arrangements. For years, platform firms have sought to obscure this reality, characterizing themselves as passive intermediaries that merely match the supply and demand of services. The Convention cuts through this narrative. By affirming the status of platform workers as bearers of rights and platform firms as bearers of corresponding duties, it situates platform work within the ambit of labor protection, regardless of how platforms define their own role.
Just as worker movements were instrumental in steering the ILO’s standard-setting process, they will remain central to the national struggles over ratification, interpretation and implementation that now follow.
The Convention’s adoption closes one chapter and opens another. Like all ILO conventions, its provisions must now be translated into national law through ratification and implementation by states. The extent to which the Convention alleviates precarity in the platform economy will depend first and foremost on the choices made by ratifying states as they transpose its provisions into domestic law and policy. At the same time, the Convention’s history serves as a reminder that legal standards acquire meaning through contestation. Just as worker movements were instrumental in steering the ILO’s standard-setting process, they will remain central to the national struggles over ratification, interpretation and implementation that now follow.
The Convention carries a salient recognition: that the algorithmic systems through which platforms organize work must be understood as mechanisms of labor governance, and therefore as legitimate subjects of regulation. Its treatment of algorithmic management thus offers a useful lens through which to assess both the protections it establishes and the questions it leaves open for national implementation. Three questions are likely to prove especially far-reaching: first, the relationship between transparency and commercial secrecy; second, the role of collective institutions in governing algorithmic systems; and third, the continued extraction and repurposing of worker-generated data. It is with these questions that the article engages.
The black box exception
Algorithmic management of work raises a basic question: what information must workers possess in order to understand and challenge the systems that govern their working lives? Convention 193 responds to this through a set of provisions on transparency, review, and dispute resolution. The practical significance of these protections will depend on how states define meaningful transparency and on the degree of deference afforded to platforms’ claims of commercial secrecy.
The Convention requires that workers be informed of the use of automated systems and the extent to which such systems affect their working conditions or access to work (Article 13). This establishes a baseline of visibility into algorithmic management. The provision does not, however, specify the content of such disclosures. It does not require information concerning, for instance, the objectives embedded within algorithmic systems, the categories of data used in decision-making, the role of human intervention within ostensibly automated processes, or the institutional procedures through which algorithmic outputs are translated into managerial decisions. Several transparency-related provisions contained in the draft Recommendation did not survive to make it into the final Convention, leaving states considerable latitude to determine what adequate disclosure entails.
The Convention establishes rights to explanation and review of certain adverse decisions generated through automated systems, including those concerning remuneration and the suspension, deactivation, or termination of worker accounts (Article 15). This offers key safeguards against some of the most consequential forms of algorithmic decision-making. The provision nonetheless leaves important issues regarding the scope of review unresolved. It applies expressly only where decisions are generated by automated systems, leaving it uncertain whether workers may challenge decisions in which automation and human discretion are intertwined. Further, while it refers to “appropriate human involvement” in review processes, it stops short of requiring substantive human review. If the provision is interpreted narrowly, it may permit procedures that merely rubber-stamp algorithmic decisions instead of subjecting them to genuine scrutiny. The right to review is also expressly guaranteed only for a narrow category of adverse decisions. As a result, core mechanisms of algorithmic management, such as task allocation, pricing, ranking systems, and performance scoring, may fall outside its scope despite exerting profound influence over workers’ livelihoods. Finally, the provision appears to link these rights to a worker’s classification, raising doubts as to whether “self-employed” or “independent” workers would be entitled to exercise these rights.
To resolve disputes, the Convention also requires states to provide workers with access to fair and effective mechanisms and remedies (Article 21). Yet, the provision is not specifically tailored to the challenges posed by automated systems, nor does it establish specialized procedures through which workers may seek explanations for, challenge, or obtain remedies for algorithmic harms. Much of the institutional architecture necessary to render these rights effective is left to national implementation.
Effective scrutiny of algorithmic systems does not depend solely on information about their design. It also requires visibility into how they operate in practice and the outcomes they produce over time.
The reach of these transparency and accountability provisions is also influenced by the Convention’s treatment of commercially sensitive information. Platforms have historically framed the debate on algorithmic transparency as a conflict between workers’ right to know and platforms’ claims to secrecy over proprietary information. The Convention requires states to protect commercially sensitive information in the process of its implementation (Article 24(4)). But what constitutes commercially sensitive information, and what forms of disclosure are regarded as incompatible with it, are themselves matters of law and policy. How states answer them will determine whether transparency rights provide workers with substantial insight into automated systems or are held at bay by expansive commercial secrecy claims.
Effective scrutiny of algorithmic systems does not depend solely on information about their design. It also requires visibility into how they operate in practice and the outcomes they produce over time. Aggregated disclosures—such as patterns of task allocation, rates of deactivation, the outcomes of review processes, the frequency with which automated decisions are reversed, and disparities across different worker groups—can reveal forms of control or discrimination that are otherwise difficult to identify purely through algorithmic parameters. For states seeking to give effect to the Convention’s transparency objectives, such disclosures demonstrate that substantial insight into algorithmic management can be achieved without necessarily implicating proprietary information.
These aspects also have implications beyond the operation of the Convention’s transparency provisions. States are required to adopt measures aimed at ensuring the correct classification of platform workers, taking into account the specificities of work performed through platforms (Article 9). In labor law, determining whether workers are genuinely independent typically turns on evidence of managerial control. It is well-established that rating systems, task-allocation mechanisms, dynamic pricing systems, performance metrics, and account-management tools may all function as mechanisms through which control is exercised. The visibility of these features to labour authorities and courts may influence how worker classification is determined.
The collective dimensions of algorithmic management
The effects of algorithmic management are rarely confined to individual workers. Platform algorithms allocate opportunities, determine prices, establish rankings, distribute incentives, and direct performance expectations across entire workforces. Their consequences are therefore often collective in nature, even when experienced individually. A change in task-allocation criteria may reduce earnings for thousands of workers simultaneously. A modification to the rating system may systematically disadvantage workers in particular locations. Performance metrics may reproduce gendered or racial hierarchies without generating a clearly discriminatory outcome for any single worker.
The Convention partially recognizes this collective reality. It requires states to respect, promote, and realize freedom of association and the effective recognition of the right to collective bargaining (Article 3(a)). It also requires that automated systems be used consistently with Fundamental Principles and Rights at Work, which include the freedom of association (Article 14). In the context of algorithmic management, information concerning automated systems must be provided not only to workers but also to their representatives and workers’ organizations (Article 13). The Convention further envisages implementation through consultation with representative organizations of workers (alongside employers) and permits implementation through collective agreements (Article 24(1)).
While worker collectives are entitled to receive information regarding automated systems under Article 13, the Convention does not establish corresponding rights to consultation and participation in decisions concerning their design, deployment, or modification.
These provisions acknowledge that platform work cannot be reduced to a series of bilateral interactions between workers and platforms. However, the Convention’s specific protections in relation to algorithmic management remain largely framed as individual rights. The rights to explanations and review under Article 15 are exercisable by individual workers. The right to dispute resolution and remedies under Article 21 is similarly individualized. While worker collectives are entitled to receive information regarding automated systems under Article 13, the Convention does not establish corresponding rights to consultation and participation in decisions concerning their design, deployment, or modification.
As a result, many of the institutional choices surrounding algorithmic management are left to national implementation. The Convention does not guarantee collective bargaining over the use of algorithmic systems. Nor does it establish mechanisms through which workers and their representatives can obtain information about the operation and effects of algorithmic systems across a workforce. The Convention is also silent on audits, inspections, worker participation in oversight processes, and other forms of co-governance, leaving their adoption to national implementation.
The Convention’s treatment of platform disciplining also raises collective concerns. While it requires states to establish safeguards against unlawful or discriminatory suspension, deactivation, and termination (Article 17), it does not specifically address the use of such measures by platforms in response to worker mobilisation or other forms of collective action. This is crucial, given documented concerns surrounding the use of such measures by platforms to discourage or retaliate against worker organising. Whether national laws extend specific protections against such practices will therefore shape the Convention’s capacity to support collective worker power.
Transparency can help workers understand how algorithmic systems affect them. Review procedures can provide recourse against particular decisions. But the asymmetries of power embedded in algorithmic management are unlikely to be transformed through individual rights alone. One of the enduring insights of labor law is that information acquires significance when workers have institutions capable of acting upon it. The long-term significance of these provisions will thus depend not only on the rights it grants to individual workers, but also on whether states create avenues for collective bargaining and worker representation, and oversight mechanisms through which workers can influence the systems that govern work.
The afterlives of workplace data
Convention 193 approaches workplace data through the lens of personal data protection. It requires states to establish safeguards concerning the processing of workers’ personal data and ensure that it is only processed for the legitimate purpose it was collected (Article 16(1)). Further, workers must have the right to access, rectify, and erase their data processed by platforms (Article 16(2)). These are material protections in a context where platform work is characterized by intensive forms of data extraction and surveillance.
Enterprises are known to use worker-generated data to develop new products and generate new revenue streams. Often, data gathered through platform work is repurposed to support the targeting of personalized financial products to vulnerable workers, creating new opportunities for value extraction.
However, many pressing issues raised by workplace data extraction concern what happens after it is collected. Data generated in the course of work is often aggregated, shared, repurposed, and monetized in ways that bear little relation to the original purpose for which it was collected. Enterprises are known to use worker-generated data to develop new products and generate new revenue streams. Often, data gathered through platform work is repurposed to support the targeting of personalized financial products to vulnerable workers, creating new opportunities for value extraction. The economic significance of workplace data is increasingly reflected in public policy as well. In several major cities, municipal authorities have acquired or are seeking access to ride-hailing journey data to support transport planning and urban governance.
In this context, the principle of purpose limitation embedded in Article 16(1) may prove significant in curtailing attempts to repurpose personal data collected in the course of work for uses unrelated to the original labor process. At the same time, its effectiveness will depend heavily on national implementation. Much turns on how narrowly legitimate purposes are defined and how strictly subsequent uses are scrutinized. Data initially collected for the management of labor may still be repurposed through expansive purpose definitions or broad consent notices.
A second issue concerns the scope of the Convention’s data protection provisions. Article 16 applies to workers’ personal data rather than workplace data more broadly. Many forms of value generated through platform work may be embedded in aggregated datasets, behavioural profiles, or other informational resources that do not necessarily constitute workers’ personal data. While the Convention seeks to protect workers as data subjects, it says relatively little about their interest in workplace data as producers of valuable data resources. It does not establish rights to data portability, nor does it specifically address situations in which worker-generated data has been passed on by platforms to third parties. As a result, many aspects concerning workplace data governance beyond the immediate work relationship remain unaddressed.
These questions are likely to become more significant as platform firms seek to leverage worker-generated data in the development of AI systems. Data extracted from workers is already being used to develop technologies that restructure or even displace the forms of labor from which that data was originally derived. The ability of Article 16 to constrain such downstream uses will therefore depend on how states operationalize purpose limitation and regulate secondary uses of workplace data. The governance of platform work must therefore respond not only to the conditions under which workplace data is collected, but also to the economic and social lives that such data acquires after it leaves the workplace.
The next phase
Convention 193 marks a watershed in international labor law. By recognizing platform work as work and extending basic labour guarantees to platform-mediated arrangements, it establishes a foundation for governing digitally mediated work. Its significance lies not only in the specific rights and protections it creates, but in its affirmation that the organization of work through digital platforms remains subject to the norms and institutions of labor regulation.
How states balance transparency and commercial secrecy, whether workers are granted robust collective oversight over algorithmic systems, and how worker-generated data is governed beyond the immediate employment relationship will all shape the Convention’s on-the-ground effects.
The issues explored in this article are best understood in that light. Rather than detracting from the significance of the Convention, they point to the principal terrains on which the governance of algorithmic management will be worked out through national implementation. How states balance transparency and commercial secrecy, whether workers are granted robust collective oversight over algorithmic systems, and how worker-generated data is governed beyond the immediate employment relationship will all shape the Convention’s on-the-ground effects. Addressing these questions will require more than formal transposition of the Convention’s provisions into domestic law. It will also require interpretive choices that give effect to the Convention’s underlying objectives in the face of new technological realities.
Equally important will be the legal and institutional conditions under which platform workers are able to organize and exercise collective voice. Laws governing the recognition and registration of trade unions, worker associations, and other forms of collective representation may therefore prove just as consequential as the Convention’s substantive protections.
Many of these issues also extend beyond labor law. Transparency obligations intersect with intellectual property and trade regimes. The governance of worker-generated data implicates debates that extend beyond data protection and into the political economy of data governance. Collective oversight of algorithmic systems intersects with emerging debates on AI regulation. Developments across these and related regulatory domains will inevitably influence the scope and significance of the Convention’s protections
The Convention establishes a floor. Whether it becomes a transformative instrument will depend on the political, legal, and institutional projects that build on it nationally.
