Artificial intelligence
What is AI?
Artificial intelligence (AI) encompasses a diverse range of technologies that enable machines to perform tasks such as recognising patterns, understanding language, generating content, as well as making predictions and supporting decisions. There is no single definition of AI that is used across all scientific and policy contexts. Concepts vary according to the technologies covered and the purpose of the definition.
In the governance field, the term ‘AI system’ is used more often and refers to the specific systems being developed or deployed. The OECD defines an AI system as ‘a machine-based system that, for explicit or implicit objectives, infers, from the input it receives, how to generate outputs such as predictions, content, recommendations, or decisions that can influence physical or virtual environments. Different AI systems vary in their levels of autonomy and adaptiveness after deployment.’ An almost identical definition of an AI system is used by the EU in its AI Act. For comparison, the African Union (AU), in its Continental AI Strategy, notes that AI ‘refers to computer systems that can simulate the processes of natural intelligence exhibited by humans where machines use technologies that enable them to learn and adapt, sense and interact, predict and recommend, reason and plan, optimise procedures and parameters, operate autonomously, be creative and extract knowledge from large amounts of data to make decisions and recommendations for the purpose of achieving a set of objectives identified by humans’.
AI also includes different methods and capabilities, from systems designed for specific tasks to models that can support a wide range of applications. Among the key concepts are:
- Machine learning – methods that enable systems to learn patterns from data and use them to make predictions or decisions.
- Deep learning – a type of machine learning that uses artificial neural networks with multiple layers to identify complex patterns, for example, in images and speech.
- Generative AI – systems that generate content, such as text, images, audio, video and computer code, in response to prompts or other inputs.
- Foundation models and general-purpose AI – foundation models are trained on broad datasets and can be adapted to different tasks. General-purpose AI models can be used for a wide range of purposes; definitions vary across governance frameworks.
- AI agents – systems that pursue specified objectives by taking sequences of actions, potentially using tools or interacting with other systems.
- Artificial general intelligence (AGI) – a contested concept generally referring to AI with broad capabilities across many tasks.
Usage of AI is currently common across different sectors such as healthcare, education, transport, manufacturing, and public administration. Its applications draw on different capabilities; for instance, natural language processing (NLP) enables computers to process human language, computer vision helps systems to analyse images and video, and some forms of robotics combine AI with machines that interact with the physical world. Other examples of applications of AI also cover speech recognition and predictive analytics.
Main policy issues and debates
On the one hand, AI offers opportunities for innovation and economic and social development. On the other hand, its development and use also bring up a range of policy challenges. Some of the main issues are covered below in brief.
Safety, security and international peace
AI can support scientific research, emergency response and cybersecurity, but at the same time, it can also introduce risks. For instance, AI can facilitate malicious cyber activity, as well as fraud and disinformation. AI applications in the physical world (e.g. in transportation) bring into focus issues related to human safety and the need to design systems that can properly react to unforeseen situations with minimal unintended consequences. AI systems that interact with external tools may pose additional security challenges if their actions are unpredictable or their permissions are insufficiently restricted.
Countries also increasingly look at AI as a field that could generate national security concerns, especially due to its applications in warfare and cyber defence and its implications for economic competitiveness. In international peace and security, AI can assist with conflict analysis, early warning and peacebuilding, but its military use creates concerns about decision-making and the deployment of (lethal) autonomous weapons. In this context, governments discuss the implications of such weapons and outline principles for their development and deployment, which include human control and compliance with international humanitarian law.
AI and human rights
AI can affect a wide range of human rights, such as privacy, freedom of expression, equality, and access to services. For instance, AI is increasingly used in recruitment, healthcare, education and law enforcement, where it can shape significant decisions about people’s rights and opportunities, such as who gets hired, how medical care is delivered, or which students are admitted to a school. Biased training data and the use of personal data can create risks of discrimination and privacy violations. Safeguards include lawful data processing, data security and impact assessments.
Ethics, transparency, and accountability
AI prompts debate about fairness, transparency and responsibility for decisions made or supported by AI systems. Some systems are difficult to interpret, making it harder to understand how outcomes are produced or challenge them. Policy approaches include transparency requirements, human oversight, clear responsibilities, and procedures for review and redress.
Economic and social impacts
AI may improve productivity, research, and service delivery, while changing jobs, skills requirements, and working conditions. Policy debates address worker monitoring, training, employment transitions, and social protection. Advanced AI development also requires substantial computing resources, data, and investment. The concentration of these resources has implications for competition, market power, and access to infrastructure.
AI inclusion, inequalities, and divides
The AI divide refers not only to the divide between users and non-users, but also to the divide between those who can build and govern AI and those who mainly consume it. Wealthier countries have a major advantage because they hold most of the world’s computing infrastructure, cloud capacity, investment, and AI talent. The World Bank found that high-income countries hosted 77% of global data-centre capacity (a key infrastructure for enabling AI) in 2025, while low-income countries hosted less than 0.1%.
Differences in skills and institutional capacity can create a gap too. Countries with fewer technical experts or weaker institutions may struggle to test AI systems, demand transparency, or enforce safeguards. This raises concerns that unsafe or poorly adapted systems could be deployed where oversight is weaker.
Bias can deepen these inequalities. UNESCO and UNICEF have highlighted risks for women, children and communities that are poorly represented in training data. Language remains a major gap, especially for Indigenous and minority languages.
The inclusion debate is therefore about who has the infrastructure, skills, data, language representation, and regulatory capacity to shape and meaningfully use AI.
Intellectual property and information integrity
AI development creates challenges regarding the use of copyrighted works in training data, the need for permissions and remuneration for creators, and rights over AI-generated outputs. Legal approaches differ across jurisdictions, and some issues remain contested.
Generative AI can also produce realistic but fabricated content, including deepfakes, which may facilitate fraud, impersonation, and attempts to mislead the public. Responses include content provenance, disclosure, detection tools, platform policies, and media literacy. No single technical measure can establish the authenticity of all content.
Environmental impacts
AI relies on computing hardware and data centres, which require electricity and cooling. Its environmental footprint varies with factors such as model design, usage, infrastructure efficiency and energy sources. It can also support environmental monitoring, climate-related applications and energy management, making it important to consider both resource consumption and potential benefits.
International AI governance
International AI governance involves not only intergovernmental conventions, resolutions, political commitments, and recommendations but also standards and multistakeholder processes. These differ in legal status and scope but support cooperation on cross-border issues.
United Nations framework
In August 2025, the General Assembly also established the Independent International Scientific Panel on AI and the Global Dialogue on AI Governance. The two mechanisms are intended to strengthen independent scientific assessment of AI and to provide a forum for international discussion on AI governance. The 2026 edition of the Global Dialogue, held in Geneva in July 2026, was intended to foster international cooperation and exchange on AI governance; its 2nd session is scheduled for New York in 2027. More broadly, the WSIS+20 outcome also addresses AI within wider digital cooperation and development efforts.
The UN Global Digital Compact (2024) includes commitments on AI governance, human rights and capacity building. In addition, the UN General Assembly has adopted resolutions on AI and sustainable development, AI capacity-building and AI in the military domain.
In addition, UN entities address AI within their respective areas of work. For example, UNESCO adopted the Recommendation on the Ethics of AI, a key international framework for ethical AI governance. Other examples include the ITU’s work on AI standards, the ILO’s research on AI and employment and its 2026 Convention on decent work in the platform economy (which addresses the use of algorithmic systems by platforms), WIPO’s work on AI and intellectual property, and WHO guidance on AI in healthcare. Within the UN Economic Commission for Europe (UNECE), work within the World Forum for Harmonisation of Vehicle Regulations (WP.29) has led to the adoption of a UN regulation on automated driving systems, which introduces safety rules that would apply across countries and a shared approach for checking that self-driving vehicles are safe before they reach the roads;
Other international initiatives and frameworks
Some example includes:
- OECD AI Principles – international principles for trustworthy AI.
- Global Partnership on AI (GPAI) – an international partnership working to advance human-centred, trustworthy AI in cooperation with the OECD.
- Council of Europe Framework Convention on AI – a legally binding treaty addressing AI in relation to human rights, democracy and the rule of law.
- G7 Hiroshima AI Process – international guiding principles and a voluntary code of conduct for advanced AI development.
- G20 endorsed its own non-binding AI Principles in 2019, drawn from the OECD principles.
- BRICS – with its 2025 Leaders’ Statement on AI governance being of particular relevance.
- UN Convention on Certain Conventional Weapons (CCW), Group of Governmental Experts on Lethal Autonomous Weapons Systems – multilateral discussions on autonomous weapons and issues including human control and compliance with international humanitarian law.
International AI summits
The series of international AI summits also shapes discussions and complements other institutional processes around AI governance.
- Bletchley AI Safety Summit (2023) focused on understanding and addressing risks posed by advanced AI.
- Seoul AI Summit (2024) advanced discussions on AI safety, innovation and inclusivity.
- Paris AI Action Summit (2025) emphasised inclusive AI development and the translation of shared principles into action.
- India AI Impact Summit (2026), which focused on AI’s real-world impact and inclusive development.
- Geneva AI Summit (2027), to focus on AI as a driver of prosperity and progress for all, and fostering trustworthy, responsible and safe use of AI.
National and regional governance approaches
Governments and regional organisations pursue different approaches to AI governance, as reflected by their diverse legal systems, development priorities, and institutional capacities. Their approaches range from binding legislation to policy strategies, voluntary guidance and regional cooperation.
The European Union (EU) has adopted the AI Act (2024), a binding legal framework that sets requirements according to the risks posed by AI systems. It covers areas such as prohibited AI practices, high-risk systems, and general-purpose AI models, with obligations introduced in phases.
Africa’s approach to AI governance is taking shape at several levels. At the continental level, the African Union’s Continental Artificial Intelligence Strategy frames AI as a tool for development, with an emphasis on building the infrastructure, skills, and local capacity needed to develop and use AI across the continent, while ensuring that its use remains safe and responsible. That same direction is visible at the sub-regional level. The East African Community (EAC) is developing its own Regional AI Strategy, focused on governance, infrastructure, skills, data, research, and investment across its member states. The Common Market for Eastern and Southern Africa (COMESA), in addition to developing its own Regional AI Strategy, is developing model policy guidelines and regulatory frameworks that member states can adapt to their own needs. Neighbouring countries are also working together; for instance, Francophone West African countries jointly developed guidelines for ethical, inclusive and responsible AI, reflecting an effort to build common principles around shared policy and capacity challenges. African countries are also participating in wider global AI governance efforts. Ten African states were among the founding members of the World Artificial Intelligence Cooperation Organisation, while six African countries joined the Global Privacy Assembly’s joint statement on AI-generated imagery and privacy. The overall direction is towards building AI governance that is rooted in Africa’s own development priorities while remaining connected to wider global standards and cooperation.
In Southeast Asia, the Association of Southeast Asian Nations (ASEAN) promotes regional alignment through its Guide on AI Governance and Ethics (2024). The guide shares practical recommendations for organisations on responsible AI development and deployment, including accountability, transparency, fairness, safety and human oversight. It has since expanded this work through the Guide on Generative AI, the ASEAN Responsible AI Roadmap 2025-2030 and the ASEAN AI Safety Network established in 2025.
In the Americas, the Organisation of American States (OAS) adopted a non-binding Inter-American Framework on Data Governance and Artificial Intelligence (MIGDIA, 2023) that provides policy guidance to help countries strengthen data governance and develop responsible, trustworthy AI approaches.
National AI approaches also vary and reflect different legal systems and policy priorities. For example, several jurisdictions have enacted binding AI-specific legislation, including South Korea, Vietnam, Taiwan, Peru, and Japan, among others. China also regulates specific AI applications through binding rules on issues such as algorithmic recommendations and generative AI services. Other jurisdictions have adopted AI strategies, plans and governance guidance, which vary in scope and legal status. The US’s 2025 AI Action Plan, the UK’s AI regulatory principles, Singapore’s Model AI Governance Framework, the UAE’s National Strategy for AI 2031 and India’s AI Governance Guidelines are only a few examples, as many other countries have adopted similar AI strategies or guidance.
AI standards and standardisation bodies
AI standards are agreed technical specifications and guidelines that help ensure AI systems are developed, assessed, and used consistently and reliably. They can support consistent practices across jurisdictions, but their legal effect varies: they may be voluntary or incorporated into binding regulatory requirements. They address areas such as:
- Risk management and AI management systems.
- Testing and evaluation of reliability, safety and performance.
- Transparency and documentation of AI systems.
- Fairness, security and robustness.
- Interoperability between systems.
Relevant standardisation bodies include the International Organisation for Standardisation (ISO), the International Electrotechnical Commission (IEC), the International Telecommunication Union (ITU), and the Institute of Electrical and Electronics Engineers (IEEE). In Europe, there are the European Committee for Standardisation (CEN) and the European Committee for Electrotechnical Standardisation (CENELEC), which work on standards in relation to the AI Act. The ISO/IEC AI standardisation committee develops standards covering areas such as AI trustworthiness, data and applications. In addition, the World Standards Cooperation brings together ISO, IEC and ITU and supports coordination among standards bodies. Its AI and Multimedia Authenticity Standards initiative explores how technical standards can support policy responses to AI-generated and manipulated media.
AI and other digital technologies and infrastructures
AI depends on and interacts with other technologies and infrastructure, such as cloud computing, semiconductors, and telecommunications networks. For example, AI models rely on computing power and data centre capacity, while AI can also optimise network performance and resource allocation. As a result, developing and deploying AI requires investment in these technologies and infrastructure. In turn, access to and control over them can influence which countries and companies have the resources to advance AI, with important implications for AI governance and infrastructure
Integration of AI is also possible in technologies such as the Internet of Things, robotics and autonomous vehicles, supporting applications from smart-city systems to automated navigation. Such intersections create technology-specific governance challenges, including securing connected devices, ensuring the safe operation of autonomous systems, or verifying the authenticity of AI-generated media.

Telecom infrastructure
AI is used to optimise network performance, conduct predictive maintenance, dynamically allocate network resources, and improve customer experience, among others.

Internet of things
The interplay between AI and IoT can be seen in multiple applications, from smart home devices and vehicle autopilot systems to drones and smart cities applications.

Semiconductors
AI algorithms are used in the design of chips, for improved performance and power efficiency, for instance. And then semiconductors themselves are used in AI hardware and research.

Quantum technologies
Although largely still a field of research, quantum computing promises enhanced computational power which, coupled with AI, can help address complex problems.

Robotics
AI is increasingly important to the development of robotic systems capable of operating in complex and changing environments.

Biotechnology
AI enables researchers to read, design, and engineer biological systems at unprecedented speed and scale, but it also lowers barriers to misuse.
What’s ahead?
Looking ahead, AI governance challenges continue to arise as the technology evolves. For instance, governments must assess the risks posed by increasingly capable AI systems, including in response to alarms around potential ‘superintelligent AI’. At the same time, states must manage competition over advanced AI models, computing power, and semiconductor supply chains, while reducing strategic dependencies on foreign technologies. These challenges also involve issues of digital sovereignty and international cooperation. How can governments monitor AI capabilities that may exceed existing oversight mechanisms? Who should be held responsible for unforeseen harms? And how can states maintain control over critical digital infrastructure and resources and, in parallel, balance national interests with the need for international cooperation? These are just some of the many questions that come to light from ongoing trends in AI.


