EXECUTIVE SUMMARY
- The rapid rise of Artificial Intelligence (AI) and its potential risks of invasion of personal privacy, violations of acute copyright and intellectual property rights, and breaches of ethical boundaries have prompted governments to prioritize setting up regulatory boundaries to manage this dynamic technological disruption.
- The momentum in addressing these aforementioned risks has accelerated over the past six months, with the EU AI Act coming close to being passed. This is the first legislation aimed at regulating AI.
- ASEAN’s response has been to advance the formulation of an ASEAN AI Guide, focusing on ethics and governance in AI applications, which is anticipated to be released in early-2024 at the Fourth ASEAN Digital Ministers’ Meeting (ADGMIN), chaired by Singapore.
- ASEAN’s approach to regulating AI is expected to be different from the EU’s impending AI Act. Any formal ASEAN-wide formulation of an AI policy will be based on ‘best practices by design’ rather than anything legally binding.
- Existing AI policy strategies found in the ASEAN region suggest three main areas of importance: 1) Development of AI capabilities for economic growth; 2) Enhancing AI-related skills and competencies, and: 3) Ethical and governance frameworks in AI applications.
* Kristina Fong Siew Leng is Lead Researcher (Economic Affairs) of the ASEAN Studies Centre, ISEAS – Yusof Ishak Institute. The author wishes to thank colleagues Dr Siwage Dharma Negara, Senior Fellow of the Regional Economic Studies Program, and Ms Joanne Lin, Lead Researcher and Co-coordinator of the ASEAN Studies Centre, for their valuable comments and suggestions.
ISEAS Perspective 2023/65, 7 August 2023
THE RUSH FOR REGULATION AS AI PROFILERATES
As the catchphrase ‘Generative Artificial Intelligence’ or ‘GenAI’ spreads around the world, catalyzed by the rapid rise in popularity of ChatGPT since its launch in November 2022 last year, governments globally are taking heed and have stepped up efforts to formulate some form of regulation, be it industry standards or, at best, guidelines to counter the potentially-adverse impact of this technology on humans and everyday life. The skew in proposed regulations in this space has adopted the notion of ‘responsible AI’, geared towards creating an AI governance framework for the greater good, and reaping the benefits of greater productivity and technological advancement without unintended consequences such as an invasion of personal privacy, violations of acute copyright and intellectual property rights, and breaches of ethical boundaries. AI without regulations can potentially place mankind in a more precarious standing against a technology that can change life as we know it in countless ways.[1]
As we have seen over the past decade, social media algorithms have changed human thinking and behaviour to the extent of influencing political participation.[2] Thus, the inherent risks of AI are glaring. However, the speed at which regulations can be established is key to managing this. For an entire economic bloc such as ASEAN, this is all the more critical. It is important to note that the intention or appetite to regulate AI is not a new construct, but it is an area that has been fraught with differing opinions on how it should be regulated, or even if it should be regulated at all. Even as the EU’s AI Act is close to being passed,[3] there exists a polarity in views amongst the Members of European Parliament (MEPs), with one main sticking point being the use of facial recognition surveillance[4] and the associated ethical issues that accompany it. The final vote in Parliament for the ban on real-time facial recognition use was 499 in favour out of 620 members. Regardless, the rest of the world closely watches the rollout of this legislation, in the belief that the Act may set the benchmark for AI-related policies around the world.[5]
That said, various countries have also started formulating their own policies for AI, including six ASEAN member states, namely Indonesia, Malaysia, the Philippines, Singapore, Thailand and Vietnam. These initiatives fall short of realizing a legally binding piece of legislation. EU- and ASEAN-wide regulations on AI will have differences due mainly to the lack of a region-wide legislative body in ASEAN, similar to the EU’s European Parliament. Thus, the result of any formal ASEAN-wide formulation of an AI policy will be based on ‘best practices by design’ rather than anything legally binding.
APPETITE FOR DIGITAL SERVICES MAY INCREASE RISKS IN AN UNREGULATED AI SPACE IN ASEAN
Should ASEAN countries be worried about AI applications and their far-reaching effects on industry and society? Is the region still far removed from the developments of other countries which are more developed in both industrial development and in the use of AI in sectors such as healthcare and manufacturing? As one indicator, based on May’s statistics of monthly visits to the ChatGPT chatbot application through both desktop and mobile means (Table 1), Indonesia was fourth in terms of global traffic share (4.7%) as compared to the US which was on top at 8.9%. It is notable that Indonesia was the only ASEAN country in the Top 5. Though ChatGPT’s reach is wide, it is not available in all countries; for example, Lao PDR and Myanmar have no access at all. Nevertheless, the application’s reach is wide, and its pattern of usage suggests that it is agnostic to levels of connectivity or the overall digital capabilities of the country. That said, in areas where overall digital connectivity may be low, this may be a case of wide geographic disparities with certain urban areas having denser digital usage.
Table 1. Traffic to the ChatGPT application in May 2023 and relative measures of digitalization
Source: similarweb database and World Bank World Development Indicators (data as at 2021)
According to UNCTAD’S digital services statistics, Indonesia’s share of digital service imports as a proportion of overall service imports is the largest in ASEAN, at 61.8% (Figure 1, left panel). The country remains a key consumer of foreign-originated digital services and this characteristic reinforces the trend seen above.[6] Singapore is a close second in this regard.
Figure 1. Digitally-deliverable services imports and exports in ASEAN (2017 vs 2021)
Source: UNCTAD (most recent data as at 2021); Note: Lao PDR data for 2021 takes the value of 2020’s data due to data unavailability.
On the other side of the coin, the Philippines’ digital services exports as a proportion of their overall services trade is the highest among ASEAN economies, with Vietnam coming in second (Figure 1, right panel). Despite a similar income-level classification by the World Bank as Lower-Middle Income economies, Vietnam is comparably more ‘connected’ than the Philippines, taking into consideration the fixed broadband and mobile subscription density in the respective countries (Table 2). All in all, this brings home two points: 1) The cross-border nature and porous borders in digital services trade have the ability to influence economies that are at differing levels of digitalization and 2) Economies that are not digitally developed (from an institutional or a regulatory perspective, especially with regards to safeguards), could face adverse risks grappling with the proliferation of disruptive technologies.
Table 2. The level of connectivity amongst ASEAN economies
Source: World Development Indicators, World Bank. Note: All data in Table 2 is at 2021 except for “Secure internet servers/1 million people” where only data up until 2020 is available.
IS ASEAN READY TO JUMP ON THE AI REGULATION BANDWAGON?
Are ASEAN economies ready for the rapid rise of AI and how important is it for these countries to have an ASEAN-wide AI policy? Given the rapid pace of AI development, it would be important at this juncture to assess where ASEAN countries are in terms of establishing or implementing their own AI-specific policies, how viable it is to have an overarching ASEAN-wide AI policy or set of guidelines, and which countries can realistically take the lead. Moreover, what will this region-wide AI policy likely look like, if it is realized?
In February 2023, ASEAN collectively agreed to establish an ASEAN ‘AI Guide’.[7] The formulation of the AI Governance and Ethics component is expected to take shape more expeditiously and may even be ready for dissemination by the end of the year or early 2024, at the Fourth ASEAN Digital Ministers’ Meeting (ADGMIN), chaired by Singapore.[8] Addressing the issue of governance and ethics at this initial stage is a promising indication of how seriously AI challenges are being taken. Individual ASEAN countries have initiated AI-related strategic policies or roadmaps. Table 3 highlights AI-related policies amongst the ASEAN-10.
Table 3. Summary of AI-related policy initiatives around ASEAN
Source: OECD. AI Policy Observatory, World Bank, Digital Economy Development Committee (Myanmar), Ministry of Science, Technology & Innovation (Malaysia), Digital Economy Council (Brunei Darussalam), Ministry of Economy and Finance (Cambodia), Department of Trade and Industry (Philippines), Smart Nation (Singapore), National Strategy On R&D and Application of Artificial Intelligence (baochinhphu.vn).
Key observations of AI strategies around the region:
1) AI strategies are at different stages of development – Only six of the ten ASEAN member states have developed their own AI strategies. Brunei, Cambodia, Lao PDR and Myanmar have not. These four have introduced national digital economy strategies but only Cambodia acknowledges AI in the text, and mostly in the context of AI industry development, except for some mention of the need for a data-driven governance system.
2) AI strategies have three common threads – For the six member states that have an AI policy, three main threads emerge: a) Development of AI as a technology for economic growth and development, b) Building up capacities, such as human capital capabilities, to reap the benefits of AI applications and c) Establishing ethical and governance frameworks for AI applications.
3) Varying degrees of international cooperation leverage – All six existing AI strategies feature international cooperation in the form of human capital development, as well as R&D collaborations. However, only two countries, Singapore and Malaysia, feature a strategy of leverage through international bodies or frameworks. In Malaysia’s case, there is the intention to tap on the ASEAN Committee on Science, Technology and Innovation (COSTI), as well as the UN Commissions on Science and Technology for Development (CSTD) to engage with international partners. Singapore’s engagement is more advanced with places in the OECD Expert Group on AI, and the European Commission’s High Level Expert Group on AI, thus giving it the ability to influence global standards.
The approach to an ASEAN-wide policy will need to take into account all of these factors when formulating AI ‘guardrails’. This is especially important in the event of rapid adoption of the EU-based laws; spillover effects on other parts of the world through companies doing business with EU-based entities can be substantial, much in the same way that the extensive General Data Protection Regulation (GDPR) affects external partners. As such, an ASEAN-wide policy must be sensitive to the wide disparity between member-state AI strategies, the different levels of digital capabilities and capacities, and the differences in institutional readiness across the region.
KEY REGULATORY ‘BUILDING BLOCKS’ FOR A BROAD-BASED AI POLICY STRATEGY
As the analysis of the respective AI strategies show, there are certain regulatory building blocks that should be integral to a robust AI policy, especially in ethics and governance. Given the key risks inherent in AI applications, such as risks of data privacy breaches, intellectual property infringement, and the creation of fake news, deepfakes and overall misinformation, it is important for regulations to be robust and holistic in coverage. Some key aspects to begin with can include data protection laws, cybersecurity regulations, intellectual property and copyright legislation, as well as consumer protection.
Table 4 below summarizes the status of these regulatory ‘building blocks’ around ASEAN.
Table 4. Summary of key pieces of legislation for a robust AI policy in ASEAN
Source: UNCTAD, World Intellectual Property Office, World Trade Organization, Intellectual Property Office (Brunei Darussalam), Department of Economic Planning and Statistics (Brunei Darussalam), MyIPO, Government of Philippines Official Gazette, Singapore Statutes Online. Note: * Brunei currently has a Data Protection Policy (2014) in place of Legislation, which is forthcoming.
At this juncture, all countries in ASEAN have (or will soon implement) some form of data privacy, cybersecurity, intellectual property and consumer protection legislation. In terms of consumer protection regulation relating to online incidents, only Brunei’s legislation does not provide for it. While copyright and intellectual property laws may protect against traditional copyright infringement cases; misinformation, misrepresentation or fake content (through means of AI technology) is an area which is still not quite developed within the legislation. In some AI policy documents, such as Singapore’s National Artificial Intelligence Strategy, there are explicit intentions to address these issues. ASEAN countries should expedite their forthcoming legislation in the areas of data privacy (Brunei and Myanmar), cyber security (Cambodia and Myanmar) and consumer protection laws (Brunei and Lao PDR) to get ASEAN countries onto a more level foundation to face a dynamic environment where AI will play a larger role in everyday life. Enacting legislation relating to distributing fake news does come with its challenges, especially when communications concerning these sorts of laws are not managed well and when enforcement is viewed as oppressive and an infringement on human rights. Malaysia’s Anti-Fake News Act (2018) had a short shelf life and was repealed only after a year.[9]
In order for ASEAN’s cooperation in AI to ward off adverse impacts of its applications, regional economies clearly need to reach a minimum threshold when it comes to key regulatory building blocks. An ASEAN Guide in AI can provide a timeline for this to materialize, and ensure safe use and application of AI, domestically and beyond. Moreover, existing ASEAN frameworks such as the ASEAN Framework on Personal Data Protection and the ASEAN Cybersecurity Cooperation Strategy, can be leveraged upon to streamline baseline targets so that they can be more expediently incorporated into the ASEAN-wide AI Guide.
BRIDGING DOMESTIC LEGISLATION TO INTERNATIONAL STANDARDS
Given the differing state of regulatory foundations and levels of digital development among ASEAN economies, the ASEAN approach to regulating AI will most likely take the form of a ‘best practices’ guide. The EU’s preference for bloc-wide AI legislation is not the only model that has been adopted. The US is also attempting a ‘best practices’ approach with their Blueprint for an AI Bill of Rights,[10] which aims to establish a framework for ‘accountable’ AI. In contrast, China has been seen to adopt a more progressive approach[11] which includes the introduction of various rules on different aspects of AI such as algorithms (2021), synthetically generated content (2022) and generative AI (2023). This dynamism mirrors the sophistication of the industry and the dynamic nature of AI.
In the ASEAN context, it will be important for a start to bridge the differences between the six ASEAN member states with AI policies with the four who do not, within the intended ASEAN AI Guide. The absence of a policy in these four countries reminds us that AI considerations may only take place after more fundamental digital capabilities are strengthened. The ASEAN-wide policy guidelines on AI may come to focus on the three common threads of AI policies found in the six member states: 1) AI capabilities development 2) development of national capacities to apply AI, and 3) creation of ethical and governance frameworks for AI use. These fundamentals address both the developmental aspect of the technology, and the risks inherent in a powerful technology if left unchecked.
Apart from broad-based strategies amongst ASEAN countries, regulatory building blocks of data protection, cybersecurity and copyrights, are almost in place, with some upcoming. That said, a more explicit AI element may need to be included in these laws, especially to account for copyright and intellectual property infringement incidents. Moreover, leveraging existing ASEAN frameworks such as the ASEAN Framework on Personal Data Protection and the ASEAN Cybersecurity Cooperation Strategy may help establish baseline targets and hence a framework for ‘best practices’ for the ASEAN AI Guide, and, more importantly, formulate actionable strategies for countries lagging behind. Getting foundational regulations up to speed is one of the more pressing goals.
Singapore will be the Chair of the ASEAN Digital Ministers’ Meeting (ADGMIN) and Related Meetings in 2024. It will be in a good position to lead ASEAN in the development of the ASEAN AI Guide on Governance and Ethics. With Singapore’s position at the European Commission’s High Level Expert Group on AI, it can guide the bloc’s AI strategy to better align it with international standards including good regulatory principles found in the EU’s AI Act. Singapore has an important role to play in bridging gaps within ASEAN on AI and at the same time, steering ASEAN’s AI guidelines toward international standards. It will be important for ASEAN to have a strategy that is not just symbolic in form but also able to safeguard the ASEAN economies in this dynamic arena.
ENDNOTES
For endnotes, please refer to the original pdf document.