As social media platforms increasingly turn to community moderation to manage content at scale, governments around the world are stepping up efforts to regulate online content through various regulatory approaches, ranging from mandating risk assessments and audits to issuing legal demands that regulate how content is governed and made accessible to users. In this evolving and complex landscape, understanding the human rights implications of community-based models has become critical.
In May 2025, the Global Network Initiative (GNI) held the first in a new series of learning calls for our members focused on the intersection of content moderation, human rights due diligence, and the human rights impacts of government demands and restrictions. This series aims to deepen collective understanding of how different models of content governance can impact the rights to freedom of expression and privacy, particularly as governments around the world become more active in regulating online spaces. As with all GNI activities, the series is held under GNI’s policies, including our antitrust and confidentiality policies.
The first call provided a foundational look at how human rights due diligence applies to the design and implementation of community moderation systems. Participants discussed the roles and responsibilities of various stakeholders, including platforms, governments, and civil society, in ensuring this due diligence. The discussion also highlighted potential government demands and restrictions that could affect rights to freedom of expression and privacy within community moderation models, and considered what human rights due diligence might look like to address these specific risks. This blog provides an overview of key takeaways from this first learning call.
Unlike traditional platform-driven content moderation, where companies set and enforce rules centrally, community moderation distributes governance – to varying extents – to users themselves. This can take many forms: decentralized rulemaking, community voting systems, or hybrid models where platform infrastructure supports but does not control content decisions. In some cases, community input is used as a weight in the algorithmic decision-making process, while in others, human moderators can directly decide on what content is shared and how.
Some communities are structured around volunteer-led governance, where policies are created and enforced from the ground up. Others operate on interest-based participation, where communities create their own rules and rely on both user feedback and voting systems to elevate or suppress content.
These models aim to support pluralistic dialogue, but they can also reflect dominant community norms and values, which may inadvertently silence minority voices. For example, moderation rules that rely heavily on written sources or formal expertise can exclude forms of knowledge not traditionally recognized by Western or academic institutions, raising concerns around epistemic injustice and cultural exclusion.
Recognizing this gap, Sarah Gilbert of Cornell University’s Citizens and Technology Lab encouraged participants to view moderation through an intersectional lens, recognizing that moderation systems, even when designed to protect freedom of expression, can replicate power imbalances. Moderation can act both as a form of oppression and resistance, depending on how it’s structured and who holds decision-making power. Applying human rights frameworks to community moderation requires not only centering freedom to speak, but also freedom from harm. This involves accounting for different levels of power: the systemic power of platforms, the social dynamics within communities, and the interpersonal relationships between users and moderation teams.
A key theme of the discussion was the need to integrate human rights due diligence (HRDD) into content governance, including with respect to decentralized systems. While some platforms retain minimal involvement in content decisions, they still bear responsibility for supporting inclusive, rights-respecting environments, especially when their tools and structures shape how communities operate.
Due diligence in this context may look different from how it applies to platforms and products with more top-down moderation models . For example, due diligence could involve examining how best to equip community moderators with tools and training, ensure transparency in decision-making, promote accessibility for underrepresented groups, and provide appropriate review mechanisms. There was also recognition that some models are evolving to be more inclusive by adopting intersectional approaches that account for differences in power and lived experience.
The discussion highlighted concerns about how government demands and legal restrictions may interact with community-led moderation. Even with regard to decentralized models, platforms’ policies and technical systems can be influenced by government pressure through legal requests or regulatory frameworks. Participants raised questions about the potential challenges these demands pose to the autonomy and privacy of users involved in community moderation, and the importance of considering human rights frameworks when responding to such requests. Transparency and accountability were noted as important principles to guide platform responses, though specific approaches are still evolving.
Some moderation models aim to reduce centralized editorial control by using collaborative labeling systems that attach notes or context to content without removing or downranking it. Importantly, participants noted that current approaches to collaborative labeling systems treat legal demands for content related to labels in the same way as other user-generated content, without special exceptions or carve-outs. These systems depend on broad user participation and seek to reduce bias by requiring consensus across diverse perspectives, even if, in most cases, applying a note or label to content does not automatically lead to its removal, demotion, or monetization penalties. These tools are still in early stages, raising important questions about transparency, accountability, and scalability.
Participants highlighted growing challenges for community moderators in identifying and evaluating AI-generated content. Concerns were raised about how AI can crowd out human contributions, complicate sourcing and verification, and potentially demotivate volunteers. One GNI company noted its ongoing work to develop clearer policies and tooling, including an upcoming Human Rights Impact Assessment on AI. Meta noted that it deploys “AI info” labels on content that has been identified as generated or modified by AI, and that contributors in its Community Notes pilot can add context to misleading AI-generated posts.
The discussion also touched on the growing role of AI-generated content and its implications for moderation. As AI becomes more widespread, distinguishing between real content becomes more complex, posing challenges for community moderators, who may be overwhelmed by the volume and nuance required to make accurate assessments.
Volunteers on platforms reliant on community moderation report mounting burdens due to the volume and complexity of AI-related content. Mistakenly flagging genuine human content as AI-generated (false positives) poses risks to trust and volunteer engagement, as well as freedom of expression and access to information
Some community-moderated platforms serve as training data sources for AI models, contributing to significant AI-generated content that can place a heavy burden on volunteer moderators. This situation raises broader concerns about the costs and impact of AI on public interest technologies, which will require further discussion moving forward.
Some platforms now label AI-generated content or require users to disclose the use of AI tools. But there is ongoing debate about how such content should be treated: Should AI content be judged differently? Should it be removed, flagged, or simply contextualized? Some platforms require AI-generated content to be clearly labeled and linked to sources, with community members reviewing these labels before they’re applied. The focus is on evaluating content based on its substance, not on whether it was created by AI, to avoid bias or false assumptions. When AI-generated content is detected, it is transparently flagged to inform users. Participants stressed the need for transparency, clarity, and consistency to build trust, both within communities and with the broader public.
As community-based moderation becomes a more prominent model for governing online spaces, ensuring that these systems uphold human rights standards will be critical. Ongoing dialogue is needed to better understand how human rights due diligence can be effectively integrated into evolving models of community moderation, particularly amid growing regulatory demands and the expanding use of AI. Upcoming sessions in this series will continue to explore the challenges, responsibilities, and practical steps for human rights due diligence on hash databases and automated filtering.