Publications & Documents

This report outlines a new methodology and provides a first exploratory analysis of technologies and applications that are at the core of recent advances in AI. Using AI-related keywords and technology classes, the study identifies AI-related patents protected in the United States in 2000-18. Among those, 'core' AI patents are selected based on their counts of AI-related forward citations. The analysis finds that, compared to other (AI and non-AI) patents, they are more original and general, and tend to be broader in technological scope. Technologies related to general AI, robotics, computer/image vision and recognition/detection are consistently listed among core AI patents, with autonomous driving and deep learning having recently become more prominent. Finally, core AI patents tend to spur innovation across AI-related domains, although some technologies – likely AI applications, such as autonomous driving or robotics – appear to increasingly contribute to developments in their own field.

This paper examines the nature and evolution of data localisation measures and their impact on business activity. It highlights that data localisation measures are growing and increasingly restrictive. By early 2023, 100 such measures were in place across 40 countries, with more than two-thirds combining local storage requirements with flow prohibition, the most restrictive form of data localisation. Insights gained from businesses operating in the e-payments, cloud computing, and air travel sectors suggest that data localisation can have unintended consequences. It not only increases operating costs, with implications for downstream users, but can also lead to increased vulnerabilities to fraud and cybersecurity risks, and reduced resilience to unexpected shocks. While international regulatory efforts have largely taken place through regional trade agreements (RTAs), this paper calls for continued monitoring of the regulatory environment with a view to informing efforts to agree on global rules that take into account legitimate public policy objectives while avoiding excessive fragmentation, especially through discussion at the WTO under the Joint Initiative on e-commerce.

The OECD AI Principles call for AI actors to be accountable for the proper functioning of their AI systems in accordance with their role, context, and ability to act. Likewise, the OECD Guidelines for Multinational Enterprises aim to minimise adverse impacts that may be associated with an enterprise’s operations, products and services. To develop ‘trustworthy’ and ‘responsible’ AI systems, there is a need to identify and manage AI risks. As calls for the development of accountability mechanisms and risk management frameworks continue to grow, interoperability would enhance efficiency and reduce enforcement and compliance costs. This report provides an analysis of the commonalities of AI risk management frameworks. It demonstrates that, while some elements may sometimes differ, all the risk management frameworks analysed follow a similar and sometimes functionally equivalent risk management process.

The digital economy is growing, with producers increasingly using digital technology to revolutionise their production processes, and with new business models being created based on the digital transformation. To improve the visibility of digitalisation in macroeconomic statistics, the Digital Supply and Use Tables (SUTs) framework has been developed under the auspices of the OECD’s Informal Advisory Group (IAG) on Measuring GDP in a Digitalised Economy. In the Digital SUTs framework, three dimensions are introduced for measuring the digital economy: the nature of the transaction (the 'how'), the goods and services produced (the 'what'), and the new digital industries (the 'who'). The OECD Handbook on Compiling Digital SUTs explains these three dimensions and includes examples. It also presents the high priority indicators that have been agreed by the IAG and includes recommended templates for producing the outputs.

Responsible science communication is crucial for fostering public trust in science and promoting evidence-based policymaking. However, in an evolving landscape shaped by the digital transformation and complex crises like the COVID-19 pandemic, science communication faces new challenges including widespread mis- and disinformation. To address these challenges, science communicators should follow key principles for responsible science communication including transparency, inclusivity, integrity, accountability, freedom and autonomy, and timeliness. Policymakers in turn are encouraged to promote these principles, invest in science communication capacity, establish crisis communication structures, support scientists in public communication, and promote scientific and digital literacy.

In 2019, the OECD Council adopted the Recommendation on Artificial Intelligence (the 'OECD AI Principles'). These include five values-based principles and five recommendations for OECD countries and adhering partner economies to promote responsible and trustworthy AI policies. This report takes stock of initiatives launched by countries worldwide to implement the OECD AI Principles which were reported to the OECD.AI Policy Observatory as of May 2023. It provides an overview of national AI strategies, including their oversight and monitoring bodies, expert advisory groups, as well as their monitoring and evaluation frameworks. It also discusses the various regulatory approaches that countries are adopting to ensure AI trustworthiness, such as ethics frameworks, AI-specific regulations, and regulatory sandboxes. Additionally, the report offers policy examples for each of the ten OECD AI Principles to facilitate cross-learning among policymakers.

Artificial intelligence (AI) offers tremendous benefits but also poses risks. Some of these risks have materialised into what are known as 'AI incidents'. Due to the widespread use of AI in various sectors, a surge in such incidents can be expected. To effectively monitor and prevent these risks, stakeholders need a precise yet adaptable definition of AI incidents. This report presents research and findings on terminology and practices related to incident definitions, encompassing both AI-specific and cross-disciplinary contexts. It establishes a knowledge base for identifying commonalities and encouraging the development of AI-specific adaptations in the future.

This report analyses the demand for positions that require skills needed to develop or work with AI systems across 14 OECD countries between 2019 and 2022. It finds that, despite rapid growth in the demand for AI skills, AI-related online vacancies comprised less than 1% of all job postings and were predominantly found in sectors such as ICT and Professional Services. Skills related to Machine Learning were the most sought after. The US-focused part of the study reveals a consistent demand for socio-emotional, foundational, and technical skills across all AI employers. However, leading firms – those who posted the most AI jobs – exhibited a higher demand for AI professionals combining technical expertise with leadership, innovation, and problem-solving skills, underscoring the importance of these competencies in the AI field.