Alliance for Universal Digital Rights

The Alliance for Universal Digital Rights (AUDRi) represents over 70 Civil society organisations from around the globe who advocate for the digital rights of women and girls. Our responses are based on inputs from our members. The primary purpose of the Global Dialogue is to provide a multistakeholder space for inclusive, evidence-based discussion on AI governance, enabling countries and organisations at different stages of digital transformation to align on norms, standards, and accountability mechanisms. This is particularly important in a rapidly evolving technological landscape where uneven development can deepen global inequalities. Its added value lies in bridging technical, legal, and societal perspectives, fostering multistakeholder collaboration across borders, and centring human rights, particularly for vulnerable populations such as women and girls affected by AI-driven harms. We believe the Global Dialogue will be successful if it is rooted in a human rights based approach that centers gender equality and addresses three critical issues. 1. Prioritise protection from harm by adopting a preventive, safety-by-design approach, embedding safeguards from the outset rather than reactive measures, and establishing robust mechanisms to both prevent and respond to AI-facilitated gender-based violence, disinformation, and exploitation. 2.. Strengthen algorithmic and platform accountability by promoting the development of uniform standards that prevent AI from amplifying gender, racial, or other social biases. Transparency and human oversight must be key components of these standards. 3. Advance regulatory harmonisation by reducing fragmentation across different national AI governance frameworks. This currently allows for regulatory arbitrage and inconsistent protections

• Safe, secure and trustworthy AI
• AI capacity-building
• Protection and promotion of human rights
• Transparency, accountability, and human oversight

A Safe, secure, and trustworthy AI is essential to increase public trust and reduce risks, including social bias, misinformation, and misuse. Without strong safeguards, AI systems can cause significant harm, particularly to vulnerable populations such as women and girls, through data exploitation, surveillance, exclusion, and online abuse. Ensuring safety requires embedding a survivor centered approach, transparency, accountability, privacy into AI design and governance. There is a need to critically examine whose safety is prioritised and ensure that these frameworks do not overlook power imbalances or disproportionately burden marginalised communities. B AI capacity-building is essential to ensure that women and girls, particularly in the Global South, are not excluded from shaping and benefiting from AI governance. Without targeted investment in skills, digital literacy, mentorship, and access, by both the government and companies, existing gender and rural inequalities risk being deepened, limiting their participation to passive users rather than active decision-makers. Inclusive and feminist approaches to capacity-building can empower girls and young women to understand, question, and influence AI systems that affect their lives. However, these efforts must be paired with strong accountability measures to ensure that increased capacity does not reinforce harmful or inequitable systems. E Protection and promotion of human rights Human rights must be at the centre of AI governance, as these systems increasingly shape decisions in areas such as migration, employment, and social entitlements. Without strong protections, AI can reinforce discrimination and exclusion, disproportionately affecting vulnerable groups, particularly women and girls, as well as migrants, activists, and journalists. Ensuring adherence to fundamental rights such as privacy, equality, and non-discrimination requires transparency, accountability, and human oversight throughout AI development and use. At the same time, gaps between Global North and Global South standards highlight the need for more consistent, inclusive frameworks to ensure AI advances gender equality and broader social justice rather than deepening existing inequalities. F Transparency, accountability, and human oversight Transparency, accountability, and human oversight are critical to building trust in AI systems and preventing harm. Because many AI systems operate as "black boxes," their decisions are often difficult to understand or challenge, increasing the risk of bias, misuse, and even digital repression. Ensuring clear explanations and maintaining human responsibility enables oversight, provides avenues for redress, and promotes responsible use in both public and private sectors. Pre-deployment gender impact assessment must be the standard along with moratoria on high-risk AI systems.

Gendered Impact of AI. AI-enabled risks such as harassment, gender-based violence, and deepfakes are often amplified by digital platforms, underscoring the need for stronger corporate responsibility and safeguards. Additionally, algorithmic bias reinforces harmful stereotypes and excludes women and gender-diverse persons from access to services and opportunities. This is compounded by the lack of gender-disaggregated data and feminist perspectives in AI design. There are issues of digital harassment that can amplify harmful content, exposing women and girls to online abuse and harassment. Thus there is a need to incorporate gender-sensitive policies to ensure fairness and protection against exploitation and discrimination in AI contexts. Governance discussions should therefore pay more attention to privacy-preserving and non-extractive forms of governance, including community-based moderation, contextual review, data minimisation, and alternatives to identity-based control. Digital colonialism and data governance. Data from the global majority is often used to train AI systems without fair compensation, consent, or local control, raising issues of exploitation, extraction, digital sovereignty and an erasure of indigenous knowledge. In the AI supply chain, data labelling and content moderation are often performed by women without safeguards for their labour rights and mental health. Many governance frameworks are well developed, but their impact is uneven, particularly for marginalised groups. In practice, protections are often difficult to access, and mechanisms for enforcement and redress may be weak or absent. The environmental impact of AI. The energy and water demands of large-scale AI systems, along with growing electronic waste, large carbon emissions and mining present sustainability challenges that intersect with climate justice, particularly in vulnerable regions and disproportionately affect women and girls. Privacy and surveillance trade-offs. Increased data collection, identity verification, and facial recognition can expose vulnerable groups such as the LGBT+ community and activists to additional risks, highlighting the importance of privacy-preserving and rights-based governance models.

Across diverse regions, a consistent picture emerges of uneven AI governance marked by gaps in regulation, capacity, and implementation, alongside significant opportunities for more inclusive and coordinated approaches. In countries such as Yemen, limited funding and deteriorating digital infrastructure constrain participation in global AI development, although open-source technologies and partnerships present pathways for progress. Similarly, across parts of Africa, including Nigeria, Kenya, Zimbabwe, and East Africa more broadly, weak or fragmented regulatory frameworks create accountability gaps, allowing harmful practices such as disinformation, bias, and data misuse to proliferate. These challenges are compounded by limited technical expertise within governments, judiciaries, and civil society, as well as low levels of AI literacy, particularly among marginalised groups such as women and girls. In regions like Europe and Latin America, including Spain and Peru, more advanced regulatory frameworks exist, but implementation often lags behind technological change. This results in persistent risks related to algorithmic bias, lack of transparency, and insufficient oversight, particularly in sensitive areas such as migration, social services, and environmental sustainability. Across Asia, governance models often fail to account for the realities of smaller, community-led platforms, where resource constraints and misaligned regulations can undermine both safety and participation, especially for vulnerable or stigmatised groups. Across all regions platform accountability remains a key and overarching governance challenge. Despite these challenges, there is broad agreement on key opportunities. Strengthening institutional capacity, investing in education and digital literacy, and fostering multistakeholder partnerships are seen as essential steps. Existing global and regional initiatives provide a strong foundation but remain fragmented, highlighting the need for greater coordination. Importantly, many responses emphasise the value of localised, rights-based, and community-driven approaches to governance, which can better reflect diverse social and cultural contexts. Overall, bridging governance gaps will require aligning global principles with local realities, ensuring that AI systems are transparent, accountable, and inclusive, while actively protecting human rights and promoting equitable development.

The Dialogue must build formal links between AI and data governance so that gender-just principles, including feminist approaches to consent, collective rights, and community data sovereignty, are embedded across both processes rather than treated separately. It must also establish cross-border mechanisms to monitor AI-related gender-based harms, while ensuring these cannot be misused for state surveillance or transnational repression against women human rights defenders, journalists, activists, and gender-diverse people. Any framework should be grounded in international human rights law and subject to independent civil society oversight. The Dialogue must resist prioritising economic and technological innovation over gender-related harms. Gender equality should be treated as a measurable and reported outcome, not merely a cross-cutting principle. Finally, the Dialogue should promote international convergence on moratoria for AI systems that pose serious human rights and gender-based risks, including those enabling TFGBV, until effective gender-responsive safeguards, accountability, and remedy mechanisms are in place.

There is strong convergence around a set of existing global, regional, and grassroots initiatives that the AI Dialogue should build upon, alongside a clear articulation of its potential added value. Key global frameworks include the UNESCO Recommendation on the Ethics of Artificial Intelligence, the OECD AI Principles, the EU AI Act, and the UN Global Digital Compact, all of which provide normative guidance on human rights, transparency, and accountability. In addition, multistakeholder platforms such as the Global Partnership on AI and initiatives like AI for Good were highlighted for fostering collaboration across governments, industry, academia, and civil society. Regionally, efforts such as the African Union AI Strategy, Latin American data protection reforms, and national AI strategies are important sources of contextualized governance experience. The Global Digital Compact has a principle addressing gender equality, the AI dialogue should expand upon this. Beyond formal policy frameworks, civil society networks, grassroots organisations, and community-led initiatives, including feminist and digital rights groups, are key. They contribute practical, lived knowledge on how AI systems affect communities. These actors often operate with limited resources but play a critical role in advancing inclusive, rights-based approaches and addressing local challenges. The added value of the AI Dialogue lies in its ability to connect these fragmented efforts into a more coherent and coordinated ecosystem. It can serve as a bridge between high-level global principles and on-the-ground realities by translating normative frameworks into actionable, context-specific guidance. Additionally, the Dialogue can incorporate human rights audits of platforms' AI policies and practices. As well as amplifying underrepresented voices, particularly from the Global South, women-led organisations, and marginalised communities, ensuring that AI governance is designed more inclusively and representative. The dialogue should facilitate knowledge sharing, capacity building, and cross-sector partnerships, helping stakeholders learn from diverse experiences and align on common standards while respecting local contexts by co-creating regional agendas and promoting multistakeholder consultations. The Sao Paulo Multistakeholder Guidelines should be used to ensure that dialogue strengthens accountability, improves implementation, and ensures that AI governance is survivor-centric, practical, and advancing gender inclusion in AI governance.

The Dialogue should establish inclusive participation structures from the outset. Civil society must have formal speaking rights in plenaries, representation on expert panels, and access to regular briefings and exchanges with Member States, consistent with established UN practices. A multistakeholder advisory group with guaranteed feminist and gender-focused representation should provide substantive input, while the AI Scientific Panel must include gender expertise and Global South representation. Its assessments should also undergo structured civil society review before publication. The Dialogue should be transparent about opportunities for engagement to ensure Global South civil society groups, particularly those affected by funding cuts, can participate meaningfully without short-notice constraints. In-person meetings should also take into account resource limitations and restrictive visa regimes that shape who is able to attend. Finally, the sequencing of the Dialogue matters. Thematic discussions should begin with the communities most affected by AI systems, rather than with technical presentations. Centering practitioners, rights defenders, and impacted communities from the outset would ground discussions in lived realities, dignity, and public interest concerns, leading to more responsive and inclusive governance outcomes.

Meaningful inclusion of marginalised communities in AI governance requires more than symbolic participation; it demands active engagement, resources, and opportunities to influence decisions. Many communities face multiple barriers to participation, including limited access to technology, low digital literacy, language differences, and a lack of awareness of any governance processes. Women and girls, particularly in the Global South, are disproportionately affected, as AI systems can reinforce existing inequalities. Rural populations, people with disabilities, indigenous communities, and speakers of less common languages are also frequently excluded, resulting in governance that fails to reflect the diversity of lived experiences. To address this, accessible participation channels such as online platforms, translated materials, and simplified documentation are essential, alongside financial and technical support to reduce barriers for participants from low-resource settings. Inclusion must also move beyond consultation toward co-design, ensuring that marginalised groups have real influence over decisions. This can include initiatives like travel support for grassroots leaders, labs for auditing AI tools for bias, and pre-dialogue workshops to surface regional perspectives. Importantly, underrepresented communities contribute unique governance knowledge, often developed under conditions of censorship, stigma, or resource constraints, which is rarely recognized in formal policy spaces. Recognising and elevating this expertise allows AI governance to become context-sensitive, culturally aware, and practical. Targeted efforts are also needed to address gender gaps, promote youth leadership, and ensure the participation of persons with disabilities, including universal design principles, accessible AI systems, and tailored capacity-building. Partnerships with trusted civil society and grassroots organisations can bridge local experiences with global policy, ensuring that governance reflects the realities of diverse populations. By intentionally creating inclusive, safe, and sustained participation pathways, AI governance can become equitable, ethical, and human-centred, responsive to the needs and knowledge of the communities most affected by these technologies.

The Dialogue should adopt a bottom-up approach by holding ground-level conversations, engaging practitioners directly, and integrating practical challenges into governance design. Developing regionally led agendas can strengthen engagement, enable cross-regional learning, and encourage investment in risk assessment and preventive measures rather than primarily reactive responses. The Dialogue should also support collaborative formats such as co-learning sessions, co-sharing initiatives, and innovation labs to foster ongoing exchange across sectors and regions. In addition, more dialogues and convenings should be hosted in and centered on the Global Majority to ensure broader and more equitable participation in shaping AI governance.

Effective AI governance relies on a combination of policies, practices, platforms, and capacity-building initiatives. At the international level, frameworks like UNESCO's Recommendation on the Ethics of AI and the OECD AI Principles guide human rights, transparency, accountability, and ethical AI development. The EU AI Act offers a risk-based regulatory model with strict requirements for high-risk systems, including human oversight and data governance, while algorithmic impact assessments in Canada and the UK help identify bias, privacy risks, and ethical concerns before deployment. Multi-stakeholder platforms such as the Partnership on AI and AI4People bring together governments, academia, industry, and civil society to share best practices, foster collaboration, and increase transparency. Open-source tools like AI Fairness 360 and frameworks such as the Data Ethics Canvas allow independent evaluation, bias detection, and wider participation. Regulatory sandboxes and frugal AI governance models further support innovation while ensuring safety and accountability. Regional initiatives are equally important. In Africa, the African Continental AI Strategy and national strategies, such as Zimbabwe's National AI Strategy (2026-2030), provide context-specific guidance. Kenya's Data Protection Act (2019) and the work of the ODPC demonstrate how rights-based frameworks can protect personal data, promote accountability, and enable redress. Community-led programs like the Resource Centre for Women and Girls (RCWG) and build digital literacy, especially for girls, young women, and vulnerable populations, empowering them to engage meaningfully in AI governance. Public-private collaborations also offer concrete solutions, for example, Mexico's voluntary agreements with platforms to tackle online gender-based violence. Together, these approaches, combining ethical guidance, enforceable regulations, capacity-building, and inclusive multi-stakeholder engagement, demonstrate how AI governance can be responsible, context-sensitive, and equitable, particularly in regions with emerging digital infrastructure.