An 'Asean way' for governance of AI
text size

An 'Asean way' for governance of AI

Listen to this article
Play
Pause
This file photo shows an obstructed message on a mobile phone and a researcher using a mask to test a biometrics scam concept. (Photo: hadrian-ifeelstock)
This file photo shows an obstructed message on a mobile phone and a researcher using a mask to test a biometrics scam concept. (Photo: hadrian-ifeelstock)

As other countries plough ahead with their signature initiatives for governing AI, the pressure is on Asean to show that it, too, has something to say.

The European Union has its EU AI Act; the US has a presidential Executive Order. In parallel with these regulatory frameworks, Asean has now launched new voluntary guidance on AI governance and ethics, aiming to facilitate innovation.

But is laying out another framework that Asean most needs? While the new document will help establish common views between member states as they set their national rules, developing concrete guidelines that have buy-in from all member states is challenging given their diverse levels of digital maturity, and the rapid proliferation of frameworks in other regions like the US, EU and China.

Instead, the "Asean way" for AI governance should perhaps look more targeted. Rather than identifying common principles, which risks reinventing the wheel, Asean should take a flexible minilateral approach that pulls together interested member-states to tackle the issues that are most relevant to them, such as addressing the potential for AI to empower scammers and cybercriminals and finding ways to draw AI talent and investment to the region.

Embracing differences

Asean released its "Guide on AI Governance and Ethics" in February 2024, centring on high-level principles like transparency and human-centricity. Given the diversity among Asean countries in digital maturity and willingness to regulate, the guide's flexibility is fitting.

Rather than prescribing a one-size-fits-all solution for organisations, it provides a toolkit of options. Where it speaks to member states and the region, its recommendations often focus on innovation, capacity-building, and investing in talent.

But this very versatility could create challenges in implementation. The 87-page document may be too weighty for an SME's first foray into AI, yet too high-level to be immediately actionable. This is all the more so given the proliferation of competing guidance from established bodies, like the OECD's AI Principles or the NIST AI Risk Management Framework.

The Asean guide may wind up seeming redundant as the technology evolves and global standards evolve with it.

Rather than trying to develop a comprehensive framework, Asean's AI governance work might be better off if member states homed in on their highest priorities. The guide suggests setting up an Asean Working Group on AI Governance, which could kickstart region-wide efforts to cooperate in these areas. But many conversations may benefit from minilateral arrangements, involving subsets of the ten Asean member states.

These minilateral arrangements should emphasise localisation, asking what Asean can bring to the table. They should also emphasise flexible involvement, encouraging member states to be involved in a more ad-hoc fashion based on the issue, as different groupings of Asean countries will be motivated to solve different problems.

Smashing scams

One priority that Asean should tackle is the impact of AI systems on scams and cybercrime. Scam operations in Southeast Asia already trap hundreds of thousands of human trafficking victims and cost billions annually, but AI could turbo-charge this problem.

Large language models make it easier to generate targeted scam messages at scale, and real-time deepfakes have already been used in a high-profile case to steal US$25 million (836.5 million baht) from a multinational company in Hong Kong. Voice cloning capabilities could open up new vectors of attack, allowing scammers to make ransom scam calls in the voice of people's loved ones.

Asean needs to pre-empt abuse of these tools, some of which are not widespread purely due to the temporary good grace of tech companies. OpenAI, for example, is withholding its voice cloning model due to risks of misuse; even fiercely pro-open-source Meta refuses to release its Voicebox model.

It may only be so long before similarly capable models become popular, and Asean should make full use of this grace period before they do. Having proposed Asean's new anti-online scams working group, Thailand could be a natural advocate for this issue.

Those countries with strong and growing digital industries, like Singapore, Malaysia, Vietnam, the Philippines, and Indonesia, could also join forces to build up the region's overall talent pool. One natural starting point would be attracting foreign AI talent, given Southeast Asia's popularity with remote workers. Many Asean countries have rolled out digital nomad visas or targeted entrepreneurs and tech workers explicitly. At a regional level, there may be lessons worth sharing or perhaps even a joint tourism or visa campaign worth investing in.

These countries could also make a combined effort to draw MNCs to invest in regional talent and infrastructure, as in Microsoft's plan to upskill 2.5 million people in AI across Asean, and Google's Asia-Pacific AI Opportunity Fund. To appeal to firms wanting to enter the Southeast Asian market, they could also develop a shared repository of business opportunities and localisation partners to reduce friction for investors.

Such efforts may be a tougher sell as member states may want to compete, not cooperate -- as they did over Taylor Swift. But given the speed of the AI race, they should consider joining forces, lest other blocs eclipse them. To make the most of AI, Asean member states must lean into their and the region's particular interests -- embracing the diversity, flexibility, and pragmatism that has become a signature of the Asean way.


Shuan Ee is a Policy and Strategy Manager and Jam Kraprayoon is a researcher at the San Francisco-based Institute of AI Policy and Strategy (IAPS).

Do you like the content of this article?
COMMENT (1)