provided by the G20 Research Group
G20 Summits |
G20 Ministerials |
G20 Analysis |
Search |
About the G20 Research Group
[English]
[Français]
[Deutsch]
[Italiano]
[Portuguesa]
[Japanese]
[Chinese]
[Korean]
[Indonesian]
|
G20 Information Centre provided by the G20 Research Group |
|
|
G20 Summits |
G20 Ministerials |
G20 Analysis |
Search |
About the G20 Research Group
|
||
AI Task Force on Artificial Intelligence, Data Governance and Innovation for Sustainable Development: Chair's Statement
Cape Town, South Africa, September 30, 2025
[pdf]
The meeting of the AI Task Force on Artificial Intelligence, Data Governance and Innovation for Sustainable Development, under South Africa’s Presidency of the G20, noted the importance of optimising the benefits of AI, whilst mitigating risks. Furthermore, the meeting recognised the need for meaningful participation of all nations in shaping a fair, inclusive, safe, secure, trustworthy, responsible, ethical, and sustainable global AI landscape for the benefit of humanity.
While acknowledging the need for global cooperation, the meeting affirmed that jurisdictions are entitled to develop, adopt and regulate AI, in accordance with international law and applicable legal frameworks.
The meeting reaffirmed its commitment to risk-based and human-centric, development-oriented, innovation-friendly AI policy and governance approaches that are consistent with human rights and applicable legal frameworks on security, privacy and protection of personal data, and intellectual property rights, whilst also encouraging multistakeholder participation.
The meeting welcomed the UNGA resolution on Seizing the Opportunities for Safe, Secure and Trustworthy Artificial Intelligence Systems for Sustainable Development A/RES/78/265; the resolution on Enhancing International Cooperation on Capacity Building of AI A/RES/78/311; and stressed the importance of fostering inclusive multi-stakeholder and multilateral cooperation and global AI governance that facilitates the meaningful participation of developing countries, recalling the Global Digital Compact. The meeting further acknowledged other regional and international initiatives, such as the ITU’s AI for Good Summit and UNESCO’s Recommendation on the Ethics of AI as well as previous G20 commitments, including the G20 AI Principles adopted in 2019.
Recognising the discussions of the Digital Economy Working Group, as reflected in the Chair’s Statement of 29 September 2025 (Annex 1), the following additional key themes were considered:
The meeting recognised that data governance is an imperative of the development of responsible, trustworthy, equitable, inclusive, and sustainable AI, and we highlight the importance of the complementarity between AI and data governance. The meeting noted the G20 high-level workshop on data governance and AI, and the Annex on Data Governance Priorities for Equitable and Sustainable AI, produced by the AU, UN ODET, and RIA, as well as the Toolkit produced by UNESCO under the South African G20 Presidency. (Annex 2)
The meeting supported, in accordance with applicable legal frameworks, the adoption of safe, secure, responsible and trustworthy AI in the public sector to enhance the delivery of public services for both citizens and businesses whilst strengthening government efficiency, whilst respecting human rights as well as promoting transparency, fairness and accountability. The meeting noted the Issue Brief on Digital Transformation: Capacity Building in the Public Sector prepared by UNESCO under the South African G20 Presidency. Effective digital governance is contingent upon civil servants developing their digital competencies alongside sector-specific expertise.
The meeting encouraged the strengthening of international cooperation, to accelerate the application of AI scientific research, innovation and development across different sectors, while ensuring the security and integrity of such research. This cooperation will be guided by the need to build shared and locally grounded capabilities, including enhanced computing infrastructure, accessible datasets, targeted training and skills development, partnership and funding mechanisms, and collaborative scientific programs.
The meeting recognised that open science, open innovation, shared datasets, and open-source software and data, subject to international obligations and national rules and regulations, made available on voluntary and mutually agreed terms, can be helpful for fostering the development of local AI ecosystems.
Integrity of information: Noting with concern the continued and growing threat of abuses to information integrity, online safety and the digital economy, some G20 members participated in the workshop on Generative AI and its evolving ability to produce high-quality deep fakes at low cost, and considered efforts to improve technical detection of misinformation and disinformation, dissemination and policy response, whilst reflecting upon the steps generative AI applications and social media platforms can take to watermark and label such content. Digital literacy and skills for online users are also important to ensure greater awareness of deepfakes, and to promote online safety, leveraging on the important role of the media in alerting the public about these issues.
The creation and availability of high-quality datasets in indigenous languages as well as local languages are essential to ensuring that AI respects diversity and delivers real value for everyone. Access should be provided in a way that respects traditional knowledge, local ownership, privacy, security, human rights and other legal rights, including intellectual property rights including copyrights, and the need for innovation, transparency and fair remuneration.
Intellectual property rights, and the cultural heritage, linguistic and traditional knowledge of Indigenous Peoples, local communities should be respected across the AI lifecycle, while also establishing mechanisms to allow for innovation for the benefit of all, taking into account cultural appropriateness, respecting privacy rights and ensuring data protection.
In acknowledging that this is the first G20 Presidency to be hosted on the African continent, this Task Force notes Africa’s efforts to advance AI initiatives. AI holds significant potential to accelerate the attainment of the AU Agenda 2063 development goals, including women’s empowerment industrialisation, and inclusive growth on the continent. The meeting therefore acknowledged the “AI for Africa Initiative” which was developed by the AU in collaboration with the South African G20 Presidency, as a voluntary platform for multilateral and multistakeholder cooperation between the G20 and the African Union. We will promote access to computing power in African countries, as well as AI talent and training, high quality and representative datasets, and infrastructure, as key building blocks for AI development and adoption in Africa.
The meeting took note of initiatives such as the ITU-organised AI for Good Impact Africa, and the role that the African Continental Free Trade Area (AfCFTA) can play in creating an ecosystem to scale AI in Africa. We encourage the development of the African AI ecosystem through voluntary contributions of technical and financial resources, and the development of Africa-centric sovereign AI capabilities, based on long-term partnerships with a focus on investment models that generate sustainable value on the continent.
The meeting welcomed the establishment of the Technology Policy Assistance Facility (TPAF) by UNESCO under the South African G20 Presidency, which can support G20 members and beyond in shaping AI policies that respect human rights and aligns with national and global public interest.
The meeting encouraged all G20 members, guest countries and invited international organisations to proactively contribute to the international discourse on the governance of AI and data, and to identify optimal responses to address related risks and realise the developmental potentials.
The meeting appreciated the contributions of all G20 members, guest countries and invited international organisations. The South African Presidency of the G20 also extends its appreciation and gratitude to all those who participated as Knowledge Partners, in particular UNESCO, UN ODET, OECD, ITU, ILO, UNDP, the African Development Bank (AfDB), Research ICT Africa, and the University of Pretoria. We further value the constructive role played by civil society and engagement groups the B20, T20 and M20 in taking up the issues related to AI, data and innovation.
This workshop along with the document on “Data Governance Priorities for equitable and sustainable AI” produced under the South African Presidency, aimed to assist countries in framing approaches for trustworthy data and AI governance.
Whilst acknowledging that data governance priorities may vary amongst countries, it is evident that data governance is everywhere central in promoting equitable access to quality inputs for AI development, while also being key to consent and compensation for rights holders. Access to high-quality, multilingual, and diverse data, in compliance with applicable legal frameworks, including on security, privacy and the protection of personal data, human rights and intellectual property rights, is crucial for ethical and effective AI model training and applications.
The Task Force discussions on data governance referenced the importance of data free flow with trust, and how international cooperation can help enhance the alignment of data governance and technical standards, where appropriate, to support interoperability, access to datasets and AI innovation.
The dialogues further encouraged technology companies to contribute to national AI development initiatives including where they extract data value. Attention was given to financing options including public private partnerships, as critical to multinational, regional and local data pools to prioritise local language datasets and culturally relevant training materials.These insights underlined the value of supporting initiatives to promote appropriate data sharing between the public and private sectors and civil society, consistent with domestic privacy laws and legal frameworks, while ensuring a level-playing field for researchers, and micro-, small- and medium-sized enterprises.
The Toolkit on data governance produced under the G20 Presidency by UNESCO unpacks data systems relate to G20 themes. It also highlights the need for agile governance to respond to challenges such as synthetic data, generative AI and agentic AI amongst other factors and recognising the key role of data protection authorities and access to information regulators.
Source: Communications and Digital Technologies, South Africa
This Information System is provided by the University of Toronto Library
and the G20 Research Group at the University of Toronto.
Please send comments to:
g20@utoronto.ca
This page was last updated
November 14, 2025
 |
 |
 |
 |
All contents copyright © 2025. University of Toronto unless otherwise stated. All rights reserved.