International Lawyers Assisting Workers (ILAW) Network; Derechos Digitales; International Trade Union Confederation (ITUC) ; FoxGlove

Focusing on the issues of labor and the future of work, we believe the Global Dialogue on AI Governance must have a formal consensus that AI systems must create decent jobs and promote respect for fundamental principles and rights at work as defined by the ILO, including freedom of association, collective bargaining and a safe and healthy workplace. The issues of labor have been systematically overlooked in the GDC, and it should be addressed beyond ILO's scope and work, so it should be included on AI governance forums such as this one. Workers and their representative organizations (trade unions) must be a part of all dialogues, stakeholder engagements, consultations or negotiations regarding AI policy. This would include collective bargaining over AI systems before deployment, transparent use of AI, the right to human oversight and guardrails against bossware and abusive surveillance. Transparency requires workers to have the right to know when AI is making decisions about their hiring, firing, or discipline. A "black box" algorithm should not be allowed in the workplace. Human oversight ensures that no significant workplace decision is made solely by an AI without a meaningful human-review appeal process. Worker data needs to be protected under privacy safeguards, preventing employers or third parties from selling, using, sharing or instrumentalizing workers data against their interest. Employers must bargain with unions or worker representatives before deploying new AI systems, and when significant changes to the AI systems are made. AI must not be deployed in ways that undermine rights, worsen job quality, encourage undue worker surveillance, lessen market competition, introduce new health and safety risks, or cause harmful labor-force disruptions, and human must be in the loop at all stages of the AI decision making processes.

• Social, economic, ethical, cultural, linguistic and technical implications of AI
• Protection and promotion of human rights
• Transparency, accountability, and human oversight

Workers are directly affected by the use and deployment of AI in their workplace that impact their fundamental human and labour rights, including the freedom of association and collective bargaining, freedom from discrimination, and the right to a safe and healthy workplace. Human rights inherently includes labour and workers' rights. For workers to effectively bargain, consent, and understand how AI will be used within their workplaces, workers require transparency, accountability and human oversight over decisions impacting their work, including ensuring that workers understand how the AI system will affect their work, their workplace, hiring/firing, promotion opportunities, and pay scales. Human oversight requires a human-review process and an appeals system that involves humans in the decision making. While AI augments some tasks, many workers' jobs will be deskilled, meaning they will become simpler and more repetitive. As a result, many workers will suffer downward pressure on wages and stalled career growth. At the same time, algorithmic management leads to companies increasingly outsourcing jobs to subcontractors further lowering wages and protections. AI innovation will increasingly operate across national borders in unregulated global labor marketplace and companies will seek to avoid paying the full cost of employment through non-standard forms of employment or misclassified employment status. As with increased unemployment, replacing full-time employment with an economy of independent contractors (often misclassified as such) will put significant social and financial pressure on both workers and their national governments to provide for the economic and social welfare of workers. Further, as companies rush to develop, train, and deploy AI systems, we are seeing the rise of labor abuses in the AI industry's opaque, international data enrichment value chains, which includes content moderators.

· First, the explicit consideration of the holistic impact of AI on the future of work and workers, not only in matters of hiring, firing, pay scale and workplace discipline, but also in its effects on workers' privacy, psychosocial well-being, violence and harassment in the workplace, discrimination, and job displacement. · Second, the explicit consideration of labor rights, fiscal, gender, racial and climate justice in relation to AI life cycle, especially its infrastructure, given the proliferation of data centers in contexts that relax environmental licensing regulations or regulations that create tax exemptions that may threaten the sustainability of social welfare policies, thus affecting differently groups that are already marginalized including women, LGBTIQ+, indigenous and other ethnic groups or populations. · Third, the discriminatory and gendered-impacts of AI systems on workers, given that AI systems are programmed with inherited bias, and often a lack of ability to adjust for diverse worker-populations in terms of physical features, accents, physical capacity, etc. Discriminatory impacts will only continue to be exacerbated unless AI systems are co-designed with cultural, racial, linguistic, ethical, and social implications in mind with those who will be bearing the costs.

'Right to explanation' laws are critical, but are too technical or center on the individual; they provide sometimes code-level transparency rather than a "plain-language". Workers often lack transparency into the training data used to evaluate them. Employers often purchase AI tools from third-party vendors. When a tool is found to be biased or in violation of labor rights, a governance gap allows employers to blame the vendor's algorithm, while vendors claim the employer "misused" the tool. This creates a loop where the worker has no clear target for legal redress because there is a lack of joint and several liability within the laws and regulations. In gig economy or high-speed logistics roles there is currently no governance standard for "real-time" human intervention. By the time a human reviews an AI-driven deactivation or fine, the worker has already lost at least a day's wages. There is a gap in specific protections for employees who override or report an AI system for being unethical or unsafe. If an AI suggests a shortcut that violates safety protocols, workers often fear they will be penalized for "low efficiency" if they choose the safer, human-led route. Governments are also not yet able to engage in pre-market testing for many AI products to ensure that they do not harm workers once deployed. Even AI companies themselves do not know the full impact their products will have on the labor market generally or on workers who must interact with them. It is vital that governments have the ability to regulate AI systems and that workers have access to justice for negligently or intentionally harmful systems.

The AI Dialogue plays an essential role in fostering cooperation if it focuses on: (i) identifying and building consensus to improve, refine, and strengthen global AI governance within a human rights framework; (ii) providing a safe space for dialogue on contested issues, where all relevant stakeholders are genuinely included and their meaningful participation and exchange are facilitated; and (iii) identifying, evidence-based gaps and fractures in existing AI governance frameworks, in order to help address and fill these voids through the activation of dialogues aimed at producing concrete outcomes. In carrying out its mission, it is crucial that the Global Dialogue creates a safe and inclusive space for interaction among different international and multilateral bodies, unions, civil society organizations, and other relevant stakeholders to develop, regulate, and propose AI governance frameworks, in order to identify opportunities for interoperable coordination (such as the G20, BRICS, G7, ITU, OECD, ILO, the European Union, among others).

The Global Dialogue can build on: · The São Paulo Guidelines (developed as part of NETmundial+10) on multistakeholder engagement in debates on AI governance and other digital technologies. The Dialogue can draw on these guidelines -based on lessons learned over the past ten years of NETmundial- to design its own methodologies that facilitate dialogue and interaction among multiple stakeholders. · The evidence and recommendations produced by the Human Rights Council on AI, as well as those reports produced to date by the Office of the High Commissioner for Human Rights (and other human rigths regional bodies, such as the African and Interamerican Comission of Human Rights), specially those reports calling for moratoriums on the deployment of AI systems incompatible with human rigths and principles. · The experience gained in multisectorial forums such as UNESCO's Global Civil Society Organizations (CSO) and Academic Network on AI Ethics and Policy; OECD Global Partnertship on AI; and the experience from the Freedom Online Coalition.

In particular, civil society, workers and their representative organizations from Global Majority countries should be specially considered when contributing to the generation of evidence-based information and case studies that help identify, for example, rights violations caused by AI, informational gaps regarding how AI systems that affect rights operate, scenarios of corporate lobbying that influence local regulatory processes, the exclusion of communities and stakeholder groups from regulatory or AI system implementation discussions, and helping intersect social, gender, racial and fiscal justice into the AI life cycle. This should help to: i) inform discussions with other interested stakeholders, and ii) be considered and consulted when drafting the outputs of the Global Dialogue, as their presence helps counterbalance the interests of industries and governments. Considering that the Draft Note on Themes and Structure has already been published, it is worth ensuring that this issue is included in the first track of the thematic discussions, and that special attention is given to the methodologies that will guide the dynamics of the Dialogue, particularly with regard to participation timelines, registration processes and the effective participation of stakeholders, the documentation and publication of their contributions, and the release of the different versions of the discussion documents submitted for consideration by the Global Dialogue. In this process, we suggest actively engaging and inviting trade unions and groups of workers from both the formal and informal economy to take part, as the Global Dialogue should, to the greatest extent possible, foster responses grounded in the lived experiences of workers.

Workers and their representative organizations are underrepresented in these global discussions on AI governance. Workers are facing direct impacts on their work, workplaces, and health as a result of AI systems. In many cases, workers are at the frontlines of the impacts AI, as well as the ones that are informing these systems but are not in the discussions nor are identified as a relevant stakeholder whose interests must be included in any governance related discussions. In order to be able to ensure diverse perspectives, workers and their representative organizations should be included from various sectors both in the formal and informal economy from the Global North and Global Majority countries. In identifying these individuals, gender, race, sexual orientation and identity, ethnicity, and migration status, should all be taken into account as these workers' have varying experiences with AI and within the digital supply chain. This diversity ensures that impacts throughout the digital supply chain are accounted for and represented.

By virtue of the GDC agreements, States must make the Global Dialogue a truly multi-stakeholder space. We suggest referring to WSIS as an experience to be replicated in terms of participation and best practices. We suggest that in the Draft Note on Themes and Structure to adopt rules to ensure clear and transparent methodologies that guarantee the meaningful participation of all stakeholders; civil society must have always an active role. Affirmative measures should be adopted in order to increase the participation of groups that have traditionally been underrepresented in AI governance debates (such as workers' groups and labor unions). Information needs to be centralized in order to guarantee open and free access to drafts and other discussion papers or documents, and to track their changes and current versions. In designing these methodologies, we recommend prioritizing and giving weight to ethnic and racialized groups, gender-diverse groups, as well as those groups of workers coming from organizations and unions from both formal and informal economy based in countries of the Global Majority, and to consider and implement financial and logistical means in order to support and sustain their active engagement. Also, we recommend considering hybrid modalities should always be offered in meetings, with proper procedures to assure equal participation from both in-person and online participants, as well as measures for accessibility in different levels. Selection or nomination processes should assure regional, gender and race diversity, but also that participation is meaningful and reflect a commitment with and connection to broader processes, discussions and movements, rather than individual motivations. Tokenist practices of presenting diversity without concrete commitments to make equitable participation possible should be avoided. Finally, in order to facilitate a broad participation of stakeholders, we recommend to host the Global Dialogue in countries respectful of migrants and civil society organizations; visa and immigration restrictions may curtail the impact of this forum if measures in this sense are not duly addressed.

· Ensuring that workers and their representative organizations are able to collectively bargain over the deployment and use of AI systems in their workplaces. · Algorithmic Impact Assessments (AIAs) should be conducted with union participation before, during and after any new system is deployed. This shifts the power dynamic from reactive (complaining after a harm occurs) to proactive (preventing the harm through design). Any AIAs should recommend, when necessary, to a moratorium when the AI system in question is incompatible with human rigths and principles. · California's SB 53 (Frontier AI Transparency): Effective Jan 1, 2026, this law mandates robust whistleblower protections for employees who report AI safety risks or ethical violations, preventing companies from using NDAs to silence concerns about algorithmic harms. · Brazil: NR-1 Implementation (May 2026): A new Normative Rule from the Ministry of Labor mandates that companies map "psychosocial risks" caused by AI. If an algorithm's "workload intensity" or "supervisory conduct" causes burnout or mental distress, the company is legally liable. · EU Directive 2024/2831 on improving working conditions in platform-based work establishes a right to information (Article 9) and explanation on the logic of AI-driven decisions concerning individual workers (Article 11), helping to address lack of transparency in platform operations. The Directive also prohibits to process some type of data (including those related to emotional or psychological state or data related to private conversations, including exchanges with trade union representatives (Article 7), protecting workers' right to privacy and freedom of association.