Democracy Reporting International on behalf of the EU System for an Enabling Environment for Civil Society (EU SEE)

The first Global Dialogue will succeed if it moves from principle to protection. Three concrete outcomes are essential. First, the Dialogue should lay the groundwork for a global framework on AI and human rights — including consideration of a UN Special Rapporteur on AI and Human Rights and on establishing human rights impact assessments as a standard prerequisite for AI deployment. Special attention must be given to AI's impact on the rights to freedom of peaceful assembly and association. These rights form the enabling environment for civil society, making all other forms of civic participation possible, and they are among the most directly threatened by current AI deployments. Across the EU SEE Programme's network, governments are using facial recognition to surveil civil society actors and deploying AI-generated disinformation to delegitimize human rights defenders. Specific cases include: - In Latin America alone, the coalition Al Sur has identified 83 facial recognition initiatives across 15 countries, with 53 currently operational. - In Peru, the Metropolitan Municipality of Lima deployed facial recognition across more than 4,300 surveillance cameras, including for monitoring protests, without consent or transparency about data storage. - None of these deployments were preceded by human rights impact assessments Second, the Dialogue should discuss binding transparency obligations for AI systems used in content moderation, and the role of AI on information integrity more broadly. Platforms should publish accuracy rates for every language in which they operate, not just aggregated global figures that mask failures in languages such as Amharic, Arabic, Urdu, and Swahili. Third, the Dialogue must launch a dedicated workstream on AI in conflict and crisis settings. AI-assisted targeting systems that automated the identification of tens of thousands of individuals for military strikes, biometric mass surveillance of displaced populations, and predictive tools used to profile activists represent an urgent governance gap that remains insufficiently addressed in existing multilateral processes . Success also requires a follow-up mechanism that tracks implementation between the 2026 Geneva meeting and the 2027 New York meeting. This mechanism could extend beyond initial meetings to ensure affected civil society communities can hold states and companies accountable.

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

Protection and promotion of human rights is our foremost priority because AI is actively being used to undermine the enabling environment for civil society. AI amplifies this pattern. Transparency, accountability, and human oversight are critical because the opacity of AI systems makes accountability impossible. Platforms' AI content moderation removes legitimate documentation of human rights abuses in non-English languages while failing to detect coordinated state-sponsored harassment campaigns, leaving affected communities with no meaningful remedy. In conflict settings, AI-assisted military targeting systems have reportedly operated with as little as 20 seconds of human review per target, dissolving the chain of accountability entirely. Social, economic, ethical, cultural, linguistic, and technical implications matter because AI systems are structurally biased against the communities they increasingly govern. AI content moderation trained predominantly on English systematically misclassifies legitimate political speech in low-resource languages. The use of bots, deepfakes, and doxxing disproportionately impacts women and LGBTQI communities, and these harms are compounded by AI systems that fail to detect such content in non-English languages. AI capacity-building is essential because countries in our network lack the regulatory infrastructure to govern AI, audit algorithmic systems, or participate meaningfully in international norm-setting processes.

Three cross-cutting issues directly affecting the enabling environment for civil society are inadequately addressed by the listed themes. 1) AI-enabled repression of civic freedoms. AI is accelerating the global decline of civic space through three interconnected mechanisms. AI-powered surveillance, including facial recognition at protests and biometric tracking of activists, chills freedom of assembly. In Latin America, governments are deploying facial recognition and predictive policing tools in ways that reinforce existing patterns of discrimination against excluded communities. AI-driven content moderation restricts freedom of expression by automating removal of civic speech while failing to detect state-backed disinformation. AI-facilitated privacy violations, from spyware enhanced by machine learning to mass surveillance of messaging apps, expose civil society actors to reprisals. These patterns are documented across all regions where the EU SEE project operates. Together they constitute a distinct governance challenge that risks dilution in general discussions of "trustworthiness." The Dialogue should explicitly name the protection of civic space and the enabling environment for civil society as a governance objective. 2) AI in armed conflict. The deployment of AI-assisted targeting systems that automated the generation of tens of thousands of human targets for military strikes, reportedly with minimal human review, raises significant governance concerns. AI-powered biometric surveillance of occupied and displaced populations, predictive tools profiling civilians as security threats, and tracking systems that locate individuals in family homes for strikes demand urgent governance attention. No listed theme adequately captures the intersection of AI, international humanitarian law, and civilian protection. 3) Corporate supply-chain accountability. Technology companies provide cloud computing, facial recognition infrastructure, and AI services to governments that use them against civil society. European-made spyware is exported to authoritarian regimes. The recently adopted UN Cybercrime Convention contains provisions that could further enable AI-powered surveillance without adequate safeguards for civil society. The Dialogue should address binding corporate human rights due diligence obligations for AI supply chains, including restrictions on exports of AI surveillance tools to contexts where they enable civic space violations.

Across the EU SEE network, AI governance gaps produce compounding harms to the enabling environment for civil society. AI-powered surveillance without oversight. In the absence of AI-specific governance frameworks, governments deploy facial recognition, predictive policing, and social media monitoring tools with no transparency, oversight, or remedy. -In El Salvador, Pegasus spyware has been used to monitor civil society organizations and journalists. -In Colombia, the government acquired Pegasus with no public transparency regarding its use or targets. -In Mexico, where the military is one of the world's largest users of spyware, used against those investigating corruption and human rights abuses. -In India, spyware has also been used against human rights defenders and journalists. -In Argentina, "cyber patrol reforms" to federal police statutes have expanded surveillance of social networks. -In Uganda and Mozambique, AI-enabled CCTV systems are concentrated in opposition strongholds. Content moderation that silences civic voices. Platform AI moderation performs catastrophically in languages spoken across our network. In Ethiopia, automated tools failed to detect hate speech in Amharic and Oromo that contributed to real-world violence, while simultaneously removing legitimate civic content. -In the Middle East, benign Arabic phrases have been algorithmically mistranslated as violent commands and flagged for removal. -Meta's 2025 decision to replace fact-checkers with community notes and loosen hate speech policies compounds these failures in regions where platforms serve as the primary civic information infrastructure. Governance gaps as opportunities. Civil society participation in AI governance is primarily limited to legislative consultation processes. Countries lack publicly accessible AI registries, binding transparency requirements, and independent oversight of deployed systems. Chile is the only country in Latin America to have created a public algorithm repository. This governance vacuum represents both a challenge and an opportunity. The Dialogue should support countries in building regulatory capacity that centers civic space protections as a core component of AI governance.

The AI Dialogue's unique value lies in its universality. It is the only forum where all 193 UN member states and diverse stakeholders shape AI governance together. This must translate into three concrete functions. First, establishing a minimum floor of protections for civic space in the AI era. Existing AI governance frameworks, including the OECD Principles, the EU AI Act, and the Council of Europe Convention, were developed primarily within the Global North. Communities most affected by AI-powered repression had minimal or no role in their design. The Dialogue should facilitate agreement on universal minimum standards. These should include mandatory human rights impact assessments before deploying AI in civic space contexts and prohibitions on AI uses that are fundamentally incompatible with human rights, such as mass biometric surveillance of protest and mandatory transparency for AI content moderation across all languages. Second, serving as an accountability mechanism. The Dialogue should require states to report on AI governance measures affecting civic space, track corporate compliance with commitments, and provide civil society with a recognized platform to present evidence of AI-related harm to fundamental freedoms. The Independent International Scientific Panel on AI should produce assessments responsive to concerns raised by affected communities, not only to the priorities of technology-producing nations. Third, bridging the AI governance divide. The Dialogue should facilitate South-South cooperation on AI governance, enable countries to learn from each other's regulatory approaches, and fund national civil society participation in AI policy processes. It should also address the cross-border dimensions of AI harm, including surveillance technology exports and platform content moderation that operates across jurisdictions, that no single national framework can govern.

The Dialogue should connect with and build upon initiatives across governance, monitoring, and civil society mobilization. The EU AI Act's prohibition on biometric mass surveillance in public spaces sets a precedent the Dialogue should discuss for universal application while critically examining how such regulations affect communities outside Europe. Importantly, the AI Act explicitly excludes military and national security uses, leaving a significant governance gap around some of the most harmful AI deployments affecting civil society in conflict contexts. The Council of Europe Framework Convention's human rights-centered approach offers a model, though its membership must expand to include affected countries. The Global Digital Compact's commitments on internet shutdowns, surveillance, and civic space provide an existing accountability baseline. The Dialogue should institutionalize linkages with existing monitoring systems to ensure its deliberations are grounded in documented realities. These include the EU SEE Early Warning and Monitoring Mechanism, Freedom House's Freedom on the Net, and the #KeepItOn coalition on internet shutdowns. Regional human rights mechanisms are already developing AI-specific jurisprudence and guidance that the Dialogue should build on. In February 2026, the IACHR Special Rapporteur for Freedom of Expression launched a consultation on the impact of AI on human rights in the Americas, examining how governments deploy facial recognition at protests, use predictive policing tools that reinforce discrimination, and acquire spyware to profile civil society. This kind of regional evidence-gathering on AI's impact on civic freedoms should inform the Dialogue's deliberations and help bridge the gap between global governance discussions and the realities documented on the ground. The Dialogue's added value lies in connecting these disparate initiatives into a coherent architecture and filling the gaps between them, particularly around AI in conflict, corporate accountability, and civic space protections. It should also explore extending the prohibition on biometric mass surveillance in public spaces beyond regional frameworks towards universal application.

Civil society organizations are positioned to contribute frontline evidence through structured impact hearings where affected communities describe how AI-powered surveillance, content moderation failures, and biometric profiling affect their work and freedoms. These hearings should carry formal weight in the Dialogue's outcomes. Civil society can also contribute shadow reports tracking state and corporate AI commitments. Governments should present national measures on AI governance affecting civic space, including regulatory frameworks, procurement policies for AI surveillance tools, and enforcement actions taken. Critically, governments should commit to meaningful engagement with civil society through consultations and participatory policy processes when developing or reforming AI-related legislation, including cybercrime, data protection, and online safety laws. Too often these laws are introduced without input from the communities most affected by them. Technology companies should present disaggregated transparency reports on AI content moderation accuracy broken down by language, government requests for AI-powered surveillance assistance, and human rights due diligence processes for AI deployments in high-risk contexts. The two-day format should allocate at least 40 percent of time to interactive sessions rather than plenary statements. We propose four formats. First, thematic deep-dives co-chaired by government and civil society representatives on AI's impact on specific freedoms, including expression, assembly, and privacy. Second, closed-door solution labs where mixed stakeholder groups develop governance proposals for concrete scenarios such as AI surveillance at protests or algorithmic removal of civic speech. Third, regional caucuses enable geographically proximate stakeholders to coordinate positions. Fourth, a dedicated session where the Scientific Panel on AI presents findings and receives direct input from affected communities.

- Human rights defenders and activists directly targeted by AI. Women human rights defenders face AI-facilitated doxxing and deepfakes, including sexual deepfake content designed to deter participation in public life. LGBTQI activists experience disproportionate removal of their content by AI moderation. Environmental defenders are targeted through AI-powered digital violence when challenging corporate interests. Communities living under AI-enhanced occupation and conflict have the most urgent stake in AI governance and the least access to multilateral forums. - Digital rights organizations from the Global South possess critical evidence about AI's impact on civic space but face financial, visa, and language barriers to meaningful participation in Geneva. - Communities whose civic expression is disproportionately censored or distorted by AI moderation systems trained on limited linguistic data should be included as stakeholders with direct expertise on algorithmic bias and its real-world consequences. Meaningful inclusion also requires practical investment. This means simultaneous interpretation in all UN languages, accessible documentation, hybrid participation, and financial support for travel.

The Dialogue should be grounded in a commitment to open and uncensored exchange. At the 2025 AI for Good Summit in Geneva, a keynote speaker was pressured to alter content addressing corporate involvement in AI-enabled harm in Palestine. This experience offers an important lesson for the Global Dialogue. Clear safeguards should ensure that participants, including researchers and civil society, can engage openly on questions of corporate and state accountability without restrictions on the substance of their contributions. Reverse panels would place affected communities at the center. Activists surveilled by AI, communities whose civic expression was removed by algorithmic moderation, and populations subjected to biometric profiling would pose questions directly to government officials and technology company representatives. This shifts power dynamics and ensures responsiveness to those bearing AI's consequences.

Several concrete practices from across regions offer models for governing AI's impact on civic space. ECOWAS Court rulings in Nigeria (2022), Togo (2020), and Senegal (2025) established that governments cannot arbitrarily restrict digital civic space through internet shutdowns or social media bans. This jurisprudence, developed through strategic litigation by civil society and regional networks, should be extended to AI-enabled censorship and surveillance. The Dialogue should examine how such regional accountability mechanisms can inform global norms. Taiwan's vTaiwan participatory platform demonstrates how deliberative technology can enable inclusive citizen input into AI governance. This approach shows that AI policy need not be elite-driven and offers a replicable model for democratizing AI governance in diverse contexts. Civil society resistance to AI-enabled surveillance is producing concrete results across the Global South. - In Brazil, organized advocacy, research, and strategic litigation is challenging facial recognition deployment. - In Peru, the digital rights organization Hiperderecho filed a complaint with the Personal Data Protection Authority over the Municipality of Lima's unlawful processing of biometric data at protests. - In Sri Lanka, the 2025 public review of the Online Safety Act illustrates how civic engagement can create openings for rights-respecting AI regulation. - In Zambia, the Law Association petitioned the Constitutional Court to challenge provisions within the recently enacted Cyber Crimes and Cyber Laws. The EU SEE Early Warning and Monitoring Mechanism provides an evidence-based model for tracking how AI and digital tools affect civic space. The mechanism combines local network member reporting with structured analysis, enabling early identification of threats and evidence-based advocacy. The Dialogue is encouraged to establish a formal channel with the EU SEE project to feed this evidence directly into its deliberations and to inform the work of the Scientific Panel on AI on civic space dimensions. These practices share a common thread. They center affected communities, provide accountability that moves beyond voluntary commitments, and demonstrate that governance from below is possible even in restrictive environments.